Skip to content

FactLot MCP server

io.github.9mp96ygnbk-alt/factlot9mp96ygnbk-altstreamable-httpv0.5.0Scanned

Cited CRE facts and deterministic calculators for agents: rent, CAM, debt sizing, IRR, parcel IDs.

67 tools: 1 do something that can't be undone, 47 change things and 19 only read.

Tools: 0 delete, 1 irreversible, 47 write, 19 read.
  • Reads private data · 3
  • Reads text others write · 2
  • Can send data out · 1
  • 14 findings

Tools listed without sign-in. Remote posture

With Anlyon

1 tool here does something that can't be undone. Govern it.

  • Anlyon lists the tools and adopts none.
  • Each tool you adopt gets approvals and receipts.
  • Your agent never holds the key.
Govern this server →

A tool your agent calls directly is not routed through Anlyon and is not governed by it. Anlyon Radar only lists and labels, and these labels are the same either way.

What the FactLot MCP server can do

The FactLot MCP server is published by the GitHub account 9mp96ygnbk-alt. It has 67 tools. It can do things that cannot be undone, such as sending, paying or publishing (fact_package_send) and create or change data (47 tools: calc_rent_schedule, calc_net_effective_rent, calc_percentage_rent, calc_cam_reconciliation and 43 more), but cannot delete data. 19 tools only read.

3 tools read private data (calc_lease_vs_own, doc_compare and fetch); 2 tools bring in text other people can write (fact_package_send and fact_package_status); 1 tool can send data out of your system (fact_package_send). On its own it has all three parts of a prompt-injection exfiltration path: private data, text others write, and a way out.

The fixed checks matched 14 findings (annotation-conflict and long-description). Each one comes with the text it matched.

Its remote endpoint lists its tools without sign-in, and its tools changed 5 times since Anlyon Radar first saw them.

Last scanned Nov 21, 2026, version 0.5.0.

Tools

Grouped by what each tool does, most consequential first. Open one for its reasons and full description; descriptions are the publisher's own words.

67 tools

Can't be undone 1Claude Code: ask
  • fact_package_sendSend a fact package to another firm for confirmation through the FactLot record, and get back a link for them.Reads text others writeCan send data out1 finding

    Why it is in this group

    • The name "fact_package_send" says "send": a word for sending, and what is sent cannot be taken back.
    • The input addresses people (input.to): what it sends cannot be taken back.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    • outboundThe name "fact_package_send" says "send": it sends or publishes something to other people.
    • outboundThe input takes a destination (input.to): what the tool writes goes there.

    Findings

    • Info

      The description is 3,722 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...tatus and its line). fact_package_status reads the answers and the reliers later."

      What this check looks for
    Description, as the server wrote it
    Send a fact package to another firm for confirmation through the FactLot record, and get back a link for them. For an expense (CAM) reconciliation: the tenant's facts (lease terms and the landlord's statement figures, each with its quote and page, from doc_verify_facts or doc_abstract_lease) go in a package from fact_package_create, the arithmetic is calc_cam_reconciliation's output as calculation, and each line is an item: accept_theirs where the sender accepts the counterparty's figure, state_ours where the lease gives a different figure and the counterparty is asked to agree or dispute it. The counterparty opens the link on a plain web page (no account needed), sees every item with both figures, the quote, page and document fingerprint, and agrees or disputes each. Every accepted or agreed item is recorded as a reliance, citing the package digest. The FactLot record keeps each fact, its quote, the page and each document's SHA-256, never the documents. A send goes to an outside party and cannot be recalled. Refuses without approved_by, the full name of the person who reviewed every item, and approval_confirmed: true. Refuses a package that fails fact_package_verify's integrity checks, and an item whose fact_ids are not in the package. Real or test: a package is sent for real only through the hosted FactLot API with the subscription's own FactLot key (paid, or an email-confirmed free key) or a connection signed in with one; a real send records reliances between the named organisations at once. Every other call is a test (no key, a free connection without a key, a guest key, or a local or self-hosted FactLot): the counterparty can still open the link and answer, but nothing is recorded as a reliance or counted in the FactLot record's measure. The answer says which happened (recorded_as, test_mode, and message with the reason). Returns the package id, digest, link_for_counterparty, status_link_for_sender (private to the sender) and what was recorded. For a tenant estoppel (kind estoppel) the items are the statements from estoppel_prepare package_items, every one state_ours: the tenant confirms each, or states its own version, on the same page; the tenant gives the signer's full name and title and signs electronically on the same page (a consent statement ticked and the name typed): the FactLot record keeps the signer, the time in UTC, the consent wording's version and a SHA-256 over the answered statements, and no IP address. relying_parties gives a lender or buyer its own view link (links_for_relying_parties), which shows the statements, the tenant's answers, the signature and the document fingerprints and lets that firm record its own reliance, one per confirmed statement and never on a disputed one, and only after the tenant has signed; where any statement is disputed, the page lists each with the tenant's own version and the firm must acknowledge them first, and the reliance stores which statements were confirmed and which disputed at that moment. The relying firm must differ from the sender and the tenant. With attestation (from audit_expense_statement, estoppel_prepare or rent_roll_certify on the hosted API) the FactLot record checks the service's signature and that the items, grades, figures and facts are the ones the service computed, and records the package as attested; a package with none is recorded as unattested, and a package with an attestation that does not verify is refused. Where the FactLot record requires it, a real package of kind expense_reconciliation, estoppel or rent_roll must be attested; test packages need not be. The answer says which (attestation.status and its line). fact_package_status reads the answers and the reliers later.

    Annotations the server declares

    • readOnlyHint: false
    • idempotentHint: false
    • openWorldHint: true
    • destructiveHint: false
    • title: Send a fact package to a counterparty for confirmation
Changes things 47Claude Code: ask
  • audit_expense_portfolioAudit many landlord operating-expense (CAM) statements at once, for a tenant with many locations or for a lease audit firm or tenant-rep broker acting for one.1 finding

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Info

      The description is 2,210 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...(report_units); this tool call is not counted, as audit_expense_statement is not."

      What this check looks for
    Description, as the server wrote it
    Audit many landlord operating-expense (CAM) statements at once, for a tenant with many locations or for a lease audit firm or tenant-rep broker acting for one. Takes up to 50 audits, each the input audit_expense_statement takes (the lease's verified facts, the statement's verified facts, optionally the prior year's lines) plus a label and optionally the landlord, and one as_of date for all. Runs audit_expense_statement on each location (the same function, not a second implementation). Returns, per location, the audit's summary (findings by class, dollars at stake by class, the dispute deadline, the checks that could not run), its findings in short form with lease and statement pages, a reference to its full result (digests of its input and result; include_results returns the whole result) and its own package_items for fact_package_send; across the portfolio, totals by finding class and by check, dollars at stake in total and by landlord, the locations ranked by dollars at stake, dispute deadlines in date order with days remaining counted from as_of, repeated issues (the same check failing at several locations, or in several statement years for one landlord) and the locations whose input failed validation, each with its issues (the other locations are still audited). Deterministic and no model: same inputs, same answer, no clock. Findings are differences against quoted terms, not legal conclusions. On the hosted API the answer carries, for each location, its own attestation (the one audit_expense_statement returns for that location's input), so each location's package is sent to its own landlord and checked on its own through fact_package_send; a service without a signing key returns none and says so. On the hosted API, callers without a Pro or Team key get the summary only (counts and totals); the one free full document of an email-verified account covers a single audit (audit_expense_statement), not a portfolio, and the answer says so. Metering: the portfolio report (report_create kind expense_audit_portfolio) counts one unit of the monthly report allowance for each location audited in full (report_units); this tool call is not counted, as audit_expense_statement is not.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Audit many expense statements at once
  • audit_expense_statementAudit a landlord's operating-expense (CAM) statement for a tenant, from verified facts only.1 finding

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Info

      The description is 2,623 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...ded for 24 hours; Pro and Team include every document with the monthly allowance."

      What this check looks for
    Description, as the server wrote it
    Audit a landlord's operating-expense (CAM) statement for a tenant, from verified facts only. Give the lease's facts (doc_abstract_lease or doc_verify_facts: pro-rata share, cap, base year or stop, admin fee, gross-up, excluded categories, capital-item treatment, audit right) and the statement's facts (each billed line, share applied, fee, gross-up, cap applied, totals, estimates, the date the tenant received it), each with its quote and page and passing fact_validate; optionally the prior year's lines. Returns findings, each with a stable code, the landlord's figure, the figure the lease yields, the dollar difference, the lease and statement fact ids with the quotes and pages, the calculation steps, a confidence class (follows_from_terms: the arithmetic follows from a quoted term and the statement's own numbers; depends_on_reading: it depends on how the lease sentence is read; question_to_ask: a question, never an overcharge) and what a person must confirm; totals; the last day to dispute or invoke the audit right with the notice form the lease requires; whether the difference crosses the lease's cost-shifting threshold; package_items ready for fact_package_send (state_ours with the landlord's figure as their_value for each finding, accept_theirs for lines with no finding); and checks_not_run with the reason (a missing lease term is never guessed). Twelve checks: management fee, cap, capital items, gross-up, pro-rata share, excluded categories, duplicates, base year or stop, arithmetic, year over year, deadlines, cost-shifting. Deterministic and no model: a finding without a verified lease fact behind it is never produced; same inputs, same findings; as_of is an input. Findings are differences against quoted terms, not legal conclusions: each finding carries its quote, for a person to review before anything is sent. On the hosted API the whole answer also carries attestation, the service's signature over the items, grades and figures that fact_package_send carries, so the FactLot record can show them as computed by FactLot and unchanged since; a service without a signing key returns none and says so. On the hosted API, callers without a Pro or Team key get the summary only (finding count, total difference, deadlines, checks not run) and the answer says what is withheld. Each free, email-verified account (an fl_free_ key) gets one whole document in total across audit_expense_statement, estoppel_prepare and rent_roll_certify: that answer says so, and the report built from the same arguments is included for 24 hours; Pro and Team include every document with the monthly allowance.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Audit a landlord's expense statement against the lease
  • calc_amortization_schedulePeriod-by-period schedule of a fixed-rate loan, monthly or quarterly: an optional interest-only period, then level payments on an amortization term that can run past maturity (ball…1 finding

    Why it is in this group

    • The name "calc_amortization_schedule" says "schedule": a word for a change.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Warn

      The server marks "calc_amortization_schedule" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

      Matched text

      annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The name "calc_amortization_schedule" says "schedule": a word for a change.

      What this check looks for
    Description, as the server wrote it
    Period-by-period schedule of a fixed-rate loan, monthly or quarterly: an optional interest-only period, then level payments on an amortization term that can run past maturity (balloon), with 30/360, actual/360, or actual/365 interest (the payment on a 30/360 basis). Rows are in cents that foot: interest is rounded each period, the final payment repays the whole balance, and the totals are the row sums. Deterministic: the same inputs always give the same result, with formula ID calc.amortization_schedule@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Loan amortization schedule
  • calc_cam_reconciliationReconcile a tenant's annual CAM charge: exclusions, occupancy gross-up of variable expenses, pro rata share, controllable-expense caps (compounding_from_base, simple_from_base, yea…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Reconcile a tenant's annual CAM charge: exclusions, occupancy gross-up of variable expenses, pro rata share, controllable-expense caps (compounding_from_base, simple_from_base, year_over_year), an admin fee (before or after the cap, inside or outside it, with per-line exclusions), base year or expense stop, partial-year occupancy (move-in/move-out prorates share, cap, and base by days), and estimates paid. Returns balance due and a full trace with order-of-operations warnings. Deterministic: the same inputs always give the same result, with formula ID calc.cam_reconciliation@2, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: CAM / operating expense reconciliation
  • calc_cam_reconciliation_batchReconcile CAM for a whole building in one call: the expense pool and gross-up once, then up to 500 tenants each with share, base year or expense stop, cap, admin fee, occupancy, es…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Reconcile CAM for a whole building in one call: the expense pool and gross-up once, then up to 500 tenants each with share, base year or expense stop, cap, admin fee, occupancy, estimates paid, and lease-specific excluded categories. Returns each tenant's statement (identical to calc_cam_reconciliation for that tenant, with its trace and warnings) and a building reconciliation: total billed against the recoverable pool, the unrecovered amount, and why (unleased or vacant share, lease exclusions, gross-up, caps, base years and stops, admin fees, zero floor, rounding). Deterministic: the same inputs always give the same result, with formula ID calc.cam_reconciliation_batch@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: CAM reconciliation for every tenant (batch)
  • calc_capex_recoveryRecoverable capital expenditure under a lease clause: amortize a capital cost over its useful life or the lease-specified period, straight-line (declining interest) or level paymen…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Recoverable capital expenditure under a lease clause: amortize a capital cost over its useful life or the lease-specified period, straight-line (declining interest) or level payment, with interest on the unamortized balance or none; optionally cap each year's recovery at demonstrated cost savings (with or without carry-forward); and allocate each year to tenants by pro rata share. Every assumption is echoed with an optional lease clause citation. Deterministic: the same inputs always give the same result, with formula ID calc.capex_recovery@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Capital expenditure recovery
  • calc_critical_datesDated list of a lease's notice deadlines and option windows, from its expiration date and options: each renewal option's notice deadline or notice window and renewal term (successi…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Dated list of a lease's notice deadlines and option windows, from its expiration date and options: each renewal option's notice deadline or notice window and renewal term (successive options count back from the term before), each termination option's last day to give notice, an expansion option deadline, and other deadlines (a stated date, or N days after a date such as a ROFO notice). Calendar months; N months before a date keeps the day of month, clamped to a shorter month's end (a warning gives the month-end reading when it is later). Deadlines on a weekend or US federal holiday (5 U.S.C. §6103, observed days included) are flagged, or rolled to the preceding or following business day when business_day_roll says so. With as_of: days to each event, passed deadlines warned, and the next deadline. Steps show each computation. Deterministic: the same inputs always give the same result, with formula ID calc.critical_dates@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Lease critical dates
  • calc_dcf_pro_formaProperty-level discounted cash flow from an in-place rent roll: each lease's rent, escalations, remaining term, and expense recovery (NNN, gross, base year), then at expiration one…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Property-level discounted cash flow from an in-place rent roll: each lease's rent, escalations, remaining term, and expense recovery (NNN, gross, base year), then at expiration one blended lease weighted by the renewal probability (renewal vs new-tenant rent, free rent, TI, leasing commissions, and new-tenant downtime, with fractional downtime and free rent kept as part months rather than rounded); market rent growth, general vacancy and credit loss, operating expenses, management fee, and capital reserves. Returns the annual cash-flow table (NOI and cash flow before debt) for the hold plus one year, the exit value at a cap rate on forward (or trailing) NOI less selling costs, the value (PV at the discount rate, end-of-year or mid-year), and with a purchase price the NPV, unlevered IRR, and equity multiple, with every lease event and convention stated. Deterministic: the same inputs always give the same result, with formula ID calc.dcf_pro_forma@2, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Property DCF from a rent roll
  • calc_debt_sizingSize the maximum loan under any combination of max LTV, min DSCR, and min debt yield for a given NOI, rate, and amortization.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Size the maximum loan under any combination of max LTV, min DSCR, and min debt yield for a given NOI, rate, and amortization. Reports the binding constraint, debt service, and resulting metrics; optionally a stress table over NOI, rate, and value shocks (max loan, binding constraint, LTV, DSCR, debt yield per scenario) and the refinance shortfall against an existing loan balance. Deterministic: the same inputs always give the same result, with formula ID calc.debt_sizing@3, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Loan sizing (LTV / DSCR / debt yield)
  • calc_direct_capValue by direct capitalization: (NOI − capital reserve) ÷ cap rate, then below-the-line deductions for lease-up costs, free-rent burn-off, and other adjustments to an as-is value (…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Value by direct capitalization: (NOI − capital reserve) ÷ cap rate, then below-the-line deductions for lease-up costs, free-rent burn-off, and other adjustments to an as-is value (optionally rounded); and the cap rate a price implies, before and after reserves and all-in after the below-the-line costs. Records the NOI basis (trailing, forward, stabilized). Deterministic: the same inputs always give the same result, with formula ID calc.direct_cap@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Direct capitalization value and implied cap rate
  • calc_equity_waterfallSplit periodic project equity cash flows between LP and GP: preferred return (compounding or simple) and return of capital pro rata to capital in either order, an optional GP catch…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Split periodic project equity cash flows between LP and GP: preferred return (compounding or simple) and return of capital pro rata to capital in either order, an optional GP catch-up, IRR-hurdle tiers with promote splits measured on the LP's IRR, and a residual split. Returns per-period LP and GP contributions and distributions (with the GP's promote) that foot to the cent, and LP, GP, and project IRRs and multiples, with the conventions to confirm against the operating agreement. Deterministic: the same inputs always give the same result, with formula ID calc.equity_waterfall@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: LP/GP equity distribution waterfall
  • calc_exchange_1031Not tax advice: arithmetic only.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Not tax advice: arithmetic only. The 45-day identification and 180-day exchange deadlines of IRC §1031(a)(3) (the exchange period ends earlier at a given tax-return due date), the three-property, 200%, and 95% identification tests of Treas. Reg. §1.1031(k)-1(c)(4), and the cash boot, mortgage boot, debt replacement, and trade-down of an exchange, with the boot-limited gain when the basis is given. Deterministic: the same inputs always give the same result, with formula ID calc.exchange_1031@2, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: 1031 exchange deadlines, identification, and boot
  • calc_hotel_metricsOccupancy, ADR, RevPAR, TRevPAR, and a USALI-style profit ladder for a period: departmental profit, GOP and margin, GOPPAR, management fees, EBITDA, FF&E reserve, and NOI (EBITDA l…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Occupancy, ADR, RevPAR, TRevPAR, and a USALI-style profit ladder for a period: departmental profit, GOP and margin, GOPPAR, management fees, EBITDA, FF&E reserve, and NOI (EBITDA less replacement reserve). Deterministic: the same inputs always give the same result, with formula ID calc.hotel_metrics@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Hotel operating metrics (USALI)
  • calc_irrPeriodic and annualized IRR of periodic cash flows, as decimals (0.085 = 8.5%), solved deterministically by scanning a 1-point grid of per-period rates and bisecting each sign chan…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Periodic and annualized IRR of periodic cash flows, as decimals (0.085 = 8.5%), solved deterministically by scanning a 1-point grid of per-period rates and bisecting each sign change; reports every root it finds when cash flows have multiple IRRs (two roots within one grid step can be missed). Errors when the flows do not change sign or no root lies between -99.99% and 1000% per period. Deterministic: the same inputs always give the same result, with formula ID calc.irr@2, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Internal rate of return
  • calc_lease_remeasurement_asc842Remeasure a lessee's lease liability and right-of-use asset under ASC 842 after a modification, lease-term reassessment, contingency resolution, or residual value guarantee change:…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Remeasure a lessee's lease liability and right-of-use asset under ASC 842 after a modification, lease-term reassessment, contingency resolution, or residual value guarantee change: new liability (PV of remaining payments), liability and ROU adjustments with any gain (ASC 842-20-35-4), and the per-period amortization table (operating: straight-line single lease cost; finance: interest plus amortization) in cents that foots: every row rolls forward exactly, totals are the row sums, and the final period trues up liability and ROU asset to zero. Warns where judgment is needed. Deterministic: the same inputs always give the same result, with formula ID calc.lease_remeasurement_asc842@2, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Lease remeasurement (ASC 842, lessee)
  • calc_net_effective_rentCompute effective rent, straight-line net effective rent, and discounted net effective rent ($/SF/yr) after free rent, TI allowance, leasing commissions, and other landlord concess…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Compute effective rent, straight-line net effective rent, and discounted net effective rent ($/SF/yr) after free rent, TI allowance, leasing commissions, and other landlord concessions. Rent is monthly in advance from the commencement day (no calendar proration), discounted at annual rate ÷ 12. schedule.unit (kw, mw, unit, acre, key, room, space) reports $/kW/yr, $/acre/yr and so on instead of $/SF/yr. Deterministic: the same inputs always give the same result, with formula ID calc.net_effective_rent@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Net effective rent (NER)
  • calc_npvNPV of periodic cash flows (t = 0, 1, 2, ...; the t = 0 flow is not discounted) at a per-period discount rate given as a decimal.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    NPV of periodic cash flows (t = 0, 1, 2, ...; the t = 0 flow is not discounted) at a per-period discount rate given as a decimal. Deterministic: the same inputs always give the same result, with formula ID calc.npv@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Net present value
  • calc_percentage_rentPercentage (overage) rent for a lease year: net sales after exclusions over a natural or fixed breakpoint, prorated for partial years, less amounts already paid.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Percentage (overage) rent for a lease year: net sales after exclusions over a natural or fixed breakpoint, prorated for partial years, less amounts already paid. Deterministic: the same inputs always give the same result, with formula ID calc.percentage_rent@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Retail percentage rent
  • calc_prepayment_and_defeasancePrepayment premium, total payoff, and cost as a share of balance for a fixed-rate amortizing or interest-only loan under yield maintenance (PV of remaining payments or rate differe…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Prepayment premium, total payoff, and cost as a share of balance for a fixed-rate amortizing or interest-only loan under yield maintenance (PV of remaining payments or rate differential, with a minimum and an explicit discounting convention), a step-down schedule, a flat percentage, or defeasance (cost of the securities portfolio that funds the remaining payments plus stated costs). Deterministic: the same inputs always give the same result, with formula ID calc.prepayment_and_defeasance@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Loan prepayment premium and defeasance
  • calc_reconcile_camReconcile FactLot's CAM reconciliation for one expense pool with another system's per-tenant results (Yardi or any other; parity is not claimed and is untested against a real expor…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Reconcile FactLot's CAM reconciliation for one expense pool with another system's per-tenant results (Yardi or any other; parity is not claimed and is untested against a real export): give the calc_cam_reconciliation_batch input plus, per tenant, what the other system reported (share, gross-up, cap limit and reduction, admin fee, base-year or stop deduction, billed amount, estimates, balance; billed_amount required) and a tolerance. Returns per tenant and line FactLot's amount, the other amount, the difference and whether it is within tolerance, tenants missing on either side, and likely causes worded as hypotheses to check: a different pro rata share (and, when every share differs by one factor, the pool's denominator), or a FactLot rerun that matches the other system with lines left to the default controllable: true set uncontrollable, the cap's base year one year off, the other cap method, the other gross-up target or no gross-up, the admin fee before or after and within or outside the cap, or no fee on uncontrollable lines; and differing estimates paid. Deterministic: the same inputs always give the same result, with formula ID calc.reconcile_cam@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Tie out CAM reconciliations against another system's results
  • calc_reconcile_dcfReconcile FactLot's DCF with another system's export (ARGUS or any other model; parity is not claimed and is untested against real ARGUS output): give the calc_dcf_pro_forma input …No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Reconcile FactLot's DCF with another system's export (ARGUS or any other model; parity is not claimed and is untested against real ARGUS output): give the calc_dcf_pro_forma input plus the other system's annual lines {label, values: [year 1, …]}, optionally its value, unlevered IRR and sale price, a tolerance (absolute and relative) and your label mapping. Labels map to FactLot lines by your mapping, else by common CRE line names (Potential Base Rent, Absorption & Turnover Vacancy, Free Rent, Expense Reimbursements, General Vacancy, Credit Loss, Effective Gross Revenue, Operating Expenses, Net Operating Income, Tenant Improvements, Leasing Commissions, Capital Reserves, Cash Flow Before Debt Service); positive deductions are negated (sign_convention). Returns per line and year FactLot's amount, the other amount, the difference and whether it is within tolerance; unmapped labels; and for each differing line a likely_cause from FactLot's documented conventions, worded as a hypothesis to check: differences only in rollover years (fractional downtime and free rent), a rate on a different base (general vacancy), only the exit year, a one-year shift, management fee inside operating expenses, a growth rate, a constant factor or amount, rounding; and for the value FactLot rerun with mid-year or end-of-year discounting, the other exit NOI basis, or no selling costs. Deterministic: the same inputs always give the same result, with formula ID calc.reconcile_dcf@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Tie out a DCF against another system's annual lines
  • calc_rent_scheduleBuild a base-rent schedule from commencement date, term, area, starting $/SF/yr, escalations (percent or fixed $/SF), and abated months.1 finding

    Why it is in this group

    • The name "calc_rent_schedule" says "schedule": a word for a change.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Warn

      The server marks "calc_rent_schedule" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

      Matched text

      annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The name "calc_rent_schedule" says "schedule": a word for a change.

      What this check looks for
    Description, as the server wrote it
    Build a base-rent schedule from commencement date, term, area, starting $/SF/yr, escalations (percent or fixed $/SF), and abated months. Lease months run from the commencement day-of-month with no calendar proration (warns when commencement is mid-month). Returns rent steps, totals, expiration date, and a calculation trace. Optional unit (kw, mw, unit, acre, key, room, space) relabels area and $/SF for data-center, ground, yard, residential, hotel, or parking leases without changing any number. Deterministic: the same inputs always give the same result, with formula ID calc.rent_schedule@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Lease rent schedule
  • calc_tx_unequal_appraisalEvidence calculation only: not an appraisal or an opinion of value, not legal or tax advice, and not a substitute for the advice of an attorney.1 finding

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Info

      The description is 2,248 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...esult, with formula ID calc.tx_unequal_appraisal@1, every step, and any warnings."

      What this check looks for
    Description, as the server wrote it
    Evidence calculation only: not an appraisal or an opinion of value, not legal or tax advice, and not a substitute for the advice of an attorney. The arithmetic of Tex. Tax Code §41.43(b)(3) and §42.26(a)(3), the subject's value against the median appraised value of a reasonable number of comparable properties appropriately adjusted. Takes 3–200 comparables chosen by the caller, each with its roll value and the caller's adjustments (percent, USD amount, or $/SF, each with a reason); optionally normalizes each comparable to value per SF of improvement or land area times the subject's area; applies percentages summed or compounded, then amounts; and returns every comparable's normalized and adjusted value, the sorted list, the median (for an even count, the mean of the two middle values), and the subject's value minus the median in USD (subject_minus_median) and as a fraction of the median (subject_minus_median_pct: 0.1554 means 15.54%, a decimal like every rate in FactLot, not 0.1554%), with the statutes cited. Warns on mixed value bases, §42.26(d) and the §23.231 limitation, comparables in another county, fewer than 5 comparables, total adjustments over 25%, a comparable identical to the subject, no adjustments at all (the median is then unadjusted), and, with normalize_by none, a comparable whose area differs from the subject's by more than 25%. Each comparable row carries flags (code and sentence, facts about the inputs, never a reason the tool drops anything): value per SF more than twice or less than half the subject's, a land-to-improvement area ratio more than twice or less than half the subject's, an area the subject has but the comparable lacks. Optional label (up to 80 characters) and address (up to 120) on the subject and each comparable are echoed in the result (subject_label, subject_address, comparables[i].label and .address) for reports. registry_tx_equity_candidates returns calc_args, its candidates in this tool's input shape, for the caller to review and edit before calling this tool. It does not choose comparables, invent adjustments, or say whether to protest. Deterministic: the same inputs always give the same result, with formula ID calc.tx_unequal_appraisal@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Texas unequal appraisal: median of adjusted comparables
  • calc_waltWeighted average remaining lease term in years as of a date (actual days ÷ 365.25), weighted by each lease's area_sf (weight_by "area") or annual_rent (weight_by "rent").No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Weighted average remaining lease term in years as of a date (actual days ÷ 365.25), weighted by each lease's area_sf (weight_by "area") or annual_rent (weight_by "rent"). Deterministic: the same inputs always give the same result, with formula ID calc.walt@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Weighted average lease term (WALT)
  • doc_abstract_leaseCurrent-state lease abstract from a lease and its amendments, with amendment precedence, as of as_of.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Current-state lease abstract from a lease and its amendments, with amendment precedence, as of as_of. Give each document's pages with the candidates you extracted (checked exactly as doc_verify_facts checks them), or the facts doc_verify_facts already returned for it (without pages only the value against its quote is rechecked, marked quote_on_page: false). Documents may come in any order: give the lease role original and each amendment role amendment with its effective_date (without a role the lease is inferred only when that is unambiguous, otherwise it is an input error; two or more amendments need effective dates). The original applies first, then amendments by effective_date; a later document's value for a predicate controls, and the earlier one moves to history and is linked through supersedes. An amendment effective after as_of is pending: its values are listed under pending and pending_facts with a warning and do not control. A rent schedule restated from a later date keeps the earlier steps before it (carried_forward). Each field gives its value, the controlling document, page, quote and fact, the superseded history, conflicts (two different values from one document, or from two documents of the same rank), pending values and open_items (a candidate for that predicate in the controlling or a later document that needs review or was rejected, so the value may not be final). FactLot only orders what you extracted: it does not find terms you missed, and an amendment that changes a term without restating it ("extended by 24 months") needs the resulting value extracted with that sentence as its quote, which the quote check will hold for review if the date is not written there.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Lease abstract with amendments
  • doc_check_use_conflictCheck a proposed use (a description, plus an optional use category or 2022 NAICS code) against a center's verified lease.exclusive_use and lease.prohibited_uses facts (from doc_ver…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Check a proposed use (a description, plus an optional use category or 2022 NAICS code) against a center's verified lease.exclusive_use and lease.prohibited_uses facts (from doc_verify_facts or doc_abstract_lease; each must pass fact_validate). Deterministic and textual: a restriction is a potential conflict when it shares a use category with the proposed use (FactLot's published use-category vocabulary, matched on keywords, the category given and NAICS codes) or shares words with it (two or more, or every significant word of the restriction). Each hit is labelled "potential conflict — needs human review" and returns the controlling quote and citation, the categories and words that matched, and the exclusive's carve-outs, flagging those that mention the proposed use. FactLot does not interpret the lease, decide whether a carve-out applies, or give legal conclusions; an empty result does not mean there is no conflict. Superseded facts are ignored.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Check a proposed use against exclusives
  • doc_predicatesLook up the lease predicates doc_verify_facts and doc_abstract_lease know (lease.expiration_date, lease.base_rent_schedule, lease.go_dark, ...) before extracting: for each, its val…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Look up the lease predicates doc_verify_facts and doc_abstract_lease know (lease.expiration_date, lease.base_rent_schedule, lease.go_dark, ...) before extracting: for each, its value shape as JSON Schema, one valid example with a quote that states it (and whether that example verifies), which parts of the value are checked against the quote and how (number, ratio, date, verbatim name, as-written words, a worded enum, a number inside free text), which count as checked only when the quote writes them out (summaries, some enums), and which always need review (true/false values and enums no wording confirms, such as a CAM cap method or a kick-out party) unless a person confirms them. Generated from the same tables the checks use. Give predicate for one, prefix (e.g. lease.renewal) for a group, or neither for all. Also returns notes on statuses, paths, human confirmation and the generic rules for any other dotted predicate. Reads no document.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Lease predicate catalog
  • doc_spread_statementSpread a T-12, operating statement, budget or rent roll summary YOU read, onto FactLot's published chart of accounts (multifamily, commercial, or hotel after the USALI summary), de…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Spread a T-12, operating statement, budget or rent roll summary YOU read, onto FactLot's published chart of accounts (multifamily, commercial, or hotel after the USALI summary), deterministically; FactLot runs no model. Send the pages as text and one line per item: {label, amount, quote, page}, optionally your category, a period column, and for a subtotal the indexes of the lines it sums; give the statement's own totals (effective_gross_income, total_operating_expenses, net_operating_income, ebitda, ...) as lines with that total as category. Each line is checked: the quote is on the page (a quote under 12 characters, such as a short row, only where it stands as whole words), the amount (as a number, any common written form; (1,200) and -1,200 read as negative) and the label are written in it; a line that fails is excluded with the reason, and a stated total that does not foot names the lines left out. The label is mapped to the chart by keywords: your category is used when it is in the chart, and an exception flags it for review when the dictionary maps the label elsewhere; an ambiguous or unknown label without a category is excluded. Vacancy, concessions, credit loss and other contra lines are always subtracted. Returns per period the sections, the computed totals (EGI, operating expenses, NOI, capital items, net cash flow; for hotels departmental profit, GOP, EBITDA and NOI after the replacement reserve), whether each total and subtotal the statement states foots (within 0.5 per line added, or your tolerance), every line with its citation (document, page, quote, span) and status, statistics (units, occupancy, ADR), and exceptions to review. Not checked: whether the agent missed lines, and meaning beyond the written amount and label.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Spread an operating statement
  • doc_verify_claimCheck one claim against a document page: {predicate, value} (anything doc_verify_facts accepts) or a short claim sentence ({text} or a bare string), plus the verbatim quote and pag…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Check one claim against a document page: {predicate, value} (anything doc_verify_facts accepts) or a short claim sentence ({text} or a bare string), plus the verbatim quote and page it rests on. Returns supported (the quote is on the page and every checked number, date and name in the claim is written in it), contradicted (the quote is on the page and states other numbers or dates of the same kind, listed in quote_states), or unverifiable (the quote is not on the page, nothing in the claim can be checked, or the quote states no value of that kind), with the evidence (document_id, page, quote, span) and each part checked. Textual, not semantic: for a sentence only its numbers and dates are compared, never its meaning, and a quote that mentions the right number for a different term still supports it.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Verify a claim against a document
  • doc_verify_factsCheck values YOU extracted from a document against its text, deterministically; FactLot runs no model.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Check values YOU extracted from a document against its text, deterministically; FactLot runs no model. Send the pages as text and one candidate per value: {predicate, value, quote, page}. Each candidate is checked in order: (1) the value's shape (Ingest predicates such as lease.expiration_date YYYY-MM-DD, lease.pro_rata_share in (0, 1], ordered lease.base_rent_schedule steps; any other dotted predicate needs a non-empty value); (2) the quote, at least 12 characters, is on the cited page, where only whitespace, line breaks, letter case and typographic quotes and dashes may differ (a word hyphenated across a line must keep its hyphen; a quote found on exactly one other page is accepted with the page corrected); (3) every number, date and party name in the value is written in the quote, in any common form: 1250000 matches "$1,250,000.00", 0.03 matches "3%" or "three percent", 2031-06-30 matches "June 30, 2031" or "6/30/2031"; text the lease states as written (a prohibited use, a carve-out, a protected use) is written in it as words, ignoring case, punctuation and line breaks; (4) a confidence you give below 0.8 holds it for review; (5) every other part of the value must be confirmed by the quote too: a true/false value, an enum or a summary not written in the quote holds the candidate for review, named in the reason. Verified means every part of the value was found in the quote. Returns verified (FactLot facts: method extracted, citing document_id, page, quote and character span, plus which value parts were checked and as what), needs_review (the quote is on the page but does not state the value: check consistency; low confidence: confidence; a part FactLot cannot confirm from the quote, such as a boolean: coverage) and rejected (wrong shape; quote missing, too short, or on several pages), each with a reason to act on. Not checked: meaning (a quote stating the number in another context passes), and whether the pages came from the file whose sha256 you give.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Verify extracted document facts
  • estoppel_preparePrepare the standard statements of a tenant estoppel certificate from verified lease facts, for a lender or buyer who wants the tenant to confirm the lease.1 finding

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Info

      The description is 2,617 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...ded for 24 hours; Pro and Team include every document with the monthly allowance."

      What this check looks for
    Description, as the server wrote it
    Prepare the standard statements of a tenant estoppel certificate from verified lease facts, for a lender or buyer who wants the tenant to confirm the lease. Give the lease's facts (doc_abstract_lease or doc_verify_facts, each with its quote and page and passing fact_validate: parties, commencement and expiration, rentable area, base rent schedule, security deposit, free rent, TI allowance, renewal, expansion, termination and purchase options and their notice deadlines), the lease and amendment documents, the premises and as_of, and optionally what the landlord's own records say (rent paid through, deposit held). Returns numbered statements: the lease and amendment list with document fingerprints, commencement, expiration, premises, current base rent and the next step (annual and monthly, worked and shown), security deposit, free rent, TI allowance, each option, each carrying the source quote and page, with a superseded sentence shown as superseded when an amendment changed it; and statements only the tenant can know (no other agreements, deposit paid, rent paid through, no rent paid more than one month ahead, landlord obligations outstanding, no other options, no landlord default, no tenant default, no assignment or sublease), marked tenant to confirm with no lease source. Flags every gap (a term the facts do not supply) and every conflict between the lease and an amendment, and withholds that statement rather than guess. Also returns key figures, totals, package_facts and package_items ready for fact_package_send kind estoppel. Deterministic and no model: same facts, same statements; as_of is an input. These are facts and arithmetic, not legal advice, and the tenant's signature, not FactLot, is what makes a certificate an estoppel. Each statement carries its quote, for a person to review before anything is sent. On the hosted API the whole answer also carries attestation, the service's signature over the items, grades and figures that fact_package_send carries, so the FactLot record can show them as computed by FactLot and unchanged since; a service without a signing key returns none and says so. On the hosted API, callers without a Pro or Team key get the summary only (counts, statement titles and status, gaps and conflicts) and the answer says what is withheld. Each free, email-verified account (an fl_free_ key) gets one whole document in total across audit_expense_statement, estoppel_prepare and rent_roll_certify: that answer says so, and the report built from the same arguments is included for 24 hours; Pro and Team include every document with the monthly allowance.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Prepare the statements of a tenant estoppel certificate
  • fact_package_createBundle FactLot facts (each must pass fact_validate) into a fact package a counterparty can verify: canonical JSON (RFC 8785) of every fact, an RFC 6962 Merkle tree over the facts, …1 finding

    Why it is in this group

    • The name "fact_package_create" says "create": a word for a change.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Warn

      The server marks "fact_package_create" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

      Matched text

      annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The name "fact_package_create" says "create": a word for a change.

      What this check looks for
    Description, as the server wrote it
    Bundle FactLot facts (each must pass fact_validate) into a fact package a counterparty can verify: canonical JSON (RFC 8785) of every fact, an RFC 6962 Merkle tree over the facts, the source documents the facts cite with their sha256 where known, and a digest over all of it. Returns the unsigned package, its digest, and signing_message, the exact text to sign. FactLot never holds private keys: to sign, the holder signs signing_message with their own Ed25519 key on their own machine (factlot.sign_fact_package in Python, signFactPackage in @factlot/tools, or fact_package_sign in the local MCP server). Deterministic: no clock time is added (pass created_at), and the same facts always give the same digest.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Create a fact package
  • fact_package_verifyVerify a fact package from fact_package_create: recompute every fact's hash, the Merkle root and the digest from the content (any changed fact, title, date or document list shows a…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Verify a fact package from fact_package_create: recompute every fact's hash, the Merkle root and the digest from the content (any changed fact, title, date or document list shows as a mismatch); check each fact against fact_validate; check the Ed25519 signature, against the public_key you pass (base64url, the signer's key you trust) or, without one, only against the key inside the package (which proves nothing about who signed); and, given the source documents' pages, check every quote is on its cited page. Returns valid plus each check's details, checks_run and checks_not_run (what was and was not checked), and values_not_checked: the parts of extracted facts' values that their quotes do not confirm (a true/false value, an enum, a summary), which do not change valid. Not checked: that the pages you pass are the original files (pass sha256 to tie them to the file hash), or what the facts mean.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Verify a fact package
  • id_buildingDeterministic FactLot ID for a building from its street address: county FIPS, ZIP code, and street line.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Deterministic FactLot ID for a building from its street address: county FIPS, ZIP code, and street line. Every condominium unit, suite, and floor at the address gets the same building ID, offline. id_space takes it for a suite, and registry_get shows the building's parcels.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: FactLot building ID
  • id_documentContent-addressed FactLot ID (SHA-256) for a document such as a lease, amendment, or estoppel.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Content-addressed FactLot ID (SHA-256) for a document such as a lease, amendment, or estoppel.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: FactLot document ID
  • id_entityDeterministic FactLot ID for a legal entity (LLC, LP, corporation, trust) from its registry jurisdiction and file number, or from its LEI (an LEI whose ISO 17442 check digits do no…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Deterministic FactLot ID for a legal entity (LLC, LP, corporation, trust) from its registry jurisdiction and file number, or from its LEI (an LEI whose ISO 17442 check digits do not match is rejected).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: FactLot legal-entity ID
  • id_parcelDeterministic FactLot ID for a parcel from state FIPS, county FIPS, and the assessor's parcel number (APN).No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Deterministic FactLot ID for a parcel from state FIPS, county FIPS, and the assessor's parcel number (APN). The same parcel always yields the same ID, offline. It serves as the join key across systems. APNs that differ only in where a separator falls ("12-34", "1-234") get different IDs; in a county with a documented fixed-width format the variants of one APN converge (NYC BBL "3-1234-56" = "3012340056").

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: FactLot parcel ID
  • id_spaceFactLot ID for a suite or unit within a building, given the building's FactLot URN (from id_building).No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    FactLot ID for a suite or unit within a building, given the building's FactLot URN (from id_building). Suite designators are normalized: "Suite 100", "Ste 100" and "#100" agree, while "1-20" and "120" stay distinct.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: FactLot space (suite) ID
  • registry_coverageThe counties (with parcel and building counts), entity registries, and crosswalk systems this FactLot Registry file holds, each with its sources and as-of dates.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    The counties (with parcel and building counts), entity registries, and crosswalk systems this FactLot Registry file holds, each with its sources and as-of dates. A miss outside these regions means not loaded, not absent. Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: What the registry holds
  • registry_owned_byCommercial parcels currently linked to a legal entity as owner of record (inferred links carry a confidence), each with the building it belongs to, plus a per-building summary so c…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Commercial parcels currently linked to a legal entity as owner of record (inferred links carry a confidence), each with the building it belongs to, plus a per-building summary so condominium units count once per building. On the hosted API, a call without a Pro or Team key gets the first three parcels, with every row of each and their buildings (the answer says how many were left out in portfolio_withheld and portfolio_note); a Pro or Team key gets every parcel. Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Parcels owned by an entity
  • registry_recorded_factsEvery recorded sale and document of a parcel (transaction.sale, transaction.deed, debt.mortgage_recorded, debt.mortgage_satisfaction_recorded), newest first in registry_get's order…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Every recorded sale and document of a parcel (transaction.sale, transaction.deed, debt.mortgage_recorded, debt.mortgage_satisfaction_recorded), newest first in registry_get's order, each a whole fact with its citation, a page at a time. registry_get keeps the newest 50 of them (on the hosted API without a Pro or Team key, the newest sale or deed only); when it holds more, its recorded_facts_truncated gives the totals, and this tool reads the rest. predicate limits the page to one of the four. Pass next_cursor back as cursor for the next page; next_cursor is null on the last. totals counts each predicate's facts. A price or amount is as recorded, zero or nominal included; the sources do not say whether a sale was at arm's length. Parties are organisations only. On the hosted API this tool needs a Pro or Team key. Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Recorded sales and documents of a parcel
  • registry_tx_equity_candidatesCandidate comparable parcels for a Texas unequal-appraisal comparison (Tax Code §41.43(b)(3)), from one tax year of the subject's appraisal district roll (the newest year whose rol…1 finding

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Info

      The description is 3,917 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...ecord facts about commercial property, each cited to the source file and record)."

      What this check looks for
    Description, as the server wrote it
    Candidate comparable parcels for a Texas unequal-appraisal comparison (Tax Code §41.43(b)(3)), from one tax year of the subject's appraisal district roll (the newest year whose roll gives the subject a value, or tax_year), with every field of the subject and the candidates read from that year's roll only: same county and tax year, the same state property category code, improvement area within area_band of the subject's, year built within year_band years when both are known; parcels sharing the subject's neighborhood code first (when the roll gives one), then its ZIP code, then the rest of the county if fewer than min_matches. Optional narrowing, off by default: same_building_class, same_local_use_code, local_use_codes. Each candidate and the subject carry their roll fields: id, apn (the district's account number), value and value_basis, value_per_sf (that value divided by the improvement area, to the cent: the roll's figures and one division, not an opinion of value), market and appraised value, improvement and land area, year built, state property category, local use code, neighborhood code, building class, situs address, tax year, and the citation of the value fact; the subject adds its value history by tax year. Separate land and improvement values are not returned. calc_args gives the same subject and candidates with exactly calc_tx_unequal_appraisal's input fields (tax_year, subject, comparables; id = the FactLot URN, label = the situs address, nulls left out, no adjustments): the candidate list under the default rule, not a selection by FactLot, for the caller to review and edit (remove, add or adjust comparables, each adjustment with a reason) before passing it to calc_tx_unequal_appraisal. Order of operations: registry_tx_equity_candidates, then the person's review, then calc_tx_unequal_appraisal with the reviewed calc_args, then report_create. selection echoes the rule, the effective parameters, the tax year and roll sources used, the files the compared values come from (compared_files: as-of date, and whether the file is dated before July 25 of the tax year, the certification date in Tax Code §26.01(a)), the subject's category, class, use code and area (subject_summary), each narrowing parameter and whether it was applied, how many parcels each step of the rule left (steps), the candidates counted by building class and by local use code with the subject's own beside them (class_counts, use_code_counts), and which tier each candidate came from; selection_rule restates the parameters used (tax year, category, area band and year band with their bounds, the subject summary, the narrowing applied, the step counts, counts per tier) with a notice that the list is the default rule's output (the default rule with the narrowing listed, when a narrowing was applied). Each candidate's citation.document_id names its entry in sources, which gives that snapshot's license, attribution, ai_use and redistribution, which apply to that candidate. Candidates under FactLot's default rule, a design choice, not a legal standard; a person decides which, if any, to use. For evidence calculations only: FactLot does not represent property owners, file protests, or advise whether to protest. Not an appraisal or an opinion of value (Tex. Occ. Code §1103.003(1)); not legal or tax advice and not a substitute for the advice of an attorney. Never returns owner names, mailing addresses, or ownership facts: ownership is outside this comparison and is not needed for it. Texas parcels only. Tarrant County is not served by the hosted registry while the publisher's terms for AI use are unclear. On the hosted API, a call without a Pro or Team key gets the current roll year only in the subject's value_history (the answer says so); the candidates are unchanged. Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Texas unequal-appraisal candidates
  • registry_tx_equity_candidates_batchregistry_tx_equity_candidates for many subjects in one call (1–50, e.g.1 finding

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Info

      The description is 2,150 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...ecord facts about commercial property, each cited to the source file and record)."

      What this check looks for
    Description, as the server wrote it
    registry_tx_equity_candidates for many subjects in one call (1–50, e.g. a consultant's client parcels): each subject's id (your label, echoed back) and its urn, or county_fips and apn, plus the rule parameters once for every subject (area_band, year_band, min_matches, limit, tax_year, same_building_class, same_local_use_code, local_use_codes). Each entry of results is that subject's answer from registry_tx_equity_candidates (subject, candidates, calc_args, selection, selection_rule, sources, warnings) or that subject's error (error, message, issues, and coverage on a miss), so one subject outside Texas, in Tarrant County, or not in the registry does not fail the others; the Texas-only and Tarrant checks apply to each subject. The selection rule and the notice appear once at the top (rule, notice) instead of in every answer; summary counts subjects, answered and errors. Candidates per subject only: this tool never runs calc_tx_unequal_appraisal and never picks comparables; a person decides, for each subject, which candidates (if any) to use, and only those go to calc_tx_unequal_appraisal (each subject's calc_args is its candidate list in that tool's input shape, for review and editing first). Each candidate's source rights are in that subject's sources. For evidence calculations only: FactLot does not represent property owners, file protests, or advise whether to protest. Not an appraisal or an opinion of value (Tex. Occ. Code §1103.003(1)); not legal or tax advice and not a substitute for the advice of an attorney. Never returns owner names, mailing addresses, or ownership facts: ownership is outside this comparison. On the hosted API each answered subject counts as one registry lookup, and subjects with an error are given back (a subject not in the registry inside loaded coverage as an answer that finds nothing, not counted up to an hourly limit and counted past it); this batch tool needs a Pro or Team key; registry_tx_equity_candidates, one subject per call, stays within the free daily lookups. Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Texas unequal-appraisal candidates (batch)
  • rent_roll_certifyCertify a commercial rent roll for a lender or buyer: check each line of the roll, given as verified facts (each value with its verbatim quote and page from the rent roll document)…1 finding

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Info

      The description is 2,452 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...ded for 24 hours; Pro and Team include every document with the monthly allowance."

      What this check looks for
    Description, as the server wrote it
    Certify a commercial rent roll for a lender or buyer: check each line of the roll, given as verified facts (each value with its verbatim quote and page from the rent roll document), against the leases' verified facts (and, where there is one, the answers of a confirmed estoppel), and grade it. Compares per line the tenant name, suite, rentable area, lease start and expiration, current base rent (per year, per month and per SF, the lease's worked from its rent schedule and area), next rent step, security deposit, free rent outstanding, renewal options and recoveries structure, wherever both the roll and a lease fact state it. Each line is graded: matches lease; matches lease and confirmed by tenant (the tenant agreed the premises, expiration and base rent statements of an estoppel); differs (both figures, the dollar difference and the quotes); no lease supplied; or not checkable, with the reason. A field no lease fact covers is not checked, never assumed. Totals: occupied and vacant area, occupancy, annual base rent as stated against as the leases yield and the difference, WALT (calc_walt, by area and by rent), rollover by year and top-tenant concentration. Flags: leases expiring within 12 months, tenants on the roll with no lease, leases with no roll line, rent on the roll above the lease, the roll's own arithmetic. Returns package_items and package_calculation for fact_package_send kind rent_roll. Deterministic, no model, no clock (as_of is an input). Facts and arithmetic, not legal, appraisal or lending advice: the certifying party's certification, not FactLot, is what the lender or buyer relies on, and FactLot checked the roll against the documents supplied only. On the hosted API the whole answer also carries attestation, the service's signature over the items, grades and figures that fact_package_send carries, so the FactLot record can show them as computed by FactLot and unchanged since; a service without a signing key returns none and says so. On the hosted API, callers without a Pro or Team key get the summary only (totals, key figures, grade counts, flags) and the answer says what is withheld. Each free, email-verified account (an fl_free_ key) gets one whole document in total across audit_expense_statement, estoppel_prepare and rent_roll_certify: that answer says so, and the report built from the same arguments is included for 24 hours; Pro and Team include every document with the monthly allowance.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Certify a rent roll against the leases, line by line
  • report_createHosted FactLot tool (api.factlot.com only).1 finding

    Why it is in this group

    • The name "report_create" says "create": a word for a change.

    Findings

    • Info

      The description is 5,561 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...onth (free: 3 a month per IP address; Pro 50; Team 1,000); input errors are free."

      What this check looks for
    Description, as the server wrote it
    Hosted FactLot tool (api.factlot.com only). Builds a client-ready document by code, with no model: tx_unequal_appraisal (calc_tx_unequal_appraisal: evidence packet PDF or adjustment-grid workbook with live formulas; send registry_tx_equity_candidates' sources as context.sources to print each cited source's as-of date, URL and rights, the rows of its subject and of the candidates used as comparables as context.candidates (value and improvement area must match the calculation) to print each property's account number, codes, land area and year built as sent, with value per SF computed by FactLot, and its selection_rule as context.selection_rule to print the candidate rule as applied), lease_abstract (doc_abstract_lease), cam_reconciliation (calc_cam_reconciliation or _batch), debt_sizing (calc_debt_sizing), asc842_schedule (calc_lease_remeasurement_asc842), compare_proposals (calc_compare_proposals), dcf_summary (calc_dcf_pro_forma), critical_dates (calc_critical_dates; also ics: a calendar with each notice window and deadline dated on its contractual day and a reminder 30 days before; dates before today are left out unless include_past is true), tx_protest_deadline (rules_tx_property_tax_protest_deadline on tax_year, notice_delivered_date?, as_of?, or the _batch rule when arguments has items: ics, one calendar with each notice's statutory filing deadline and conservative file-by date as all-day events, reminders 30 and 7 days before the file-by date, the rule's basis, status, warnings and notes in each description, and the rule's next_season dates when the requested year has passed; or pdf, a deadline sheet), expense_audit (audit_expense_statement: a PDF of a tenant's audit of a landlord's expense statement, each finding with its lease quote, page and calculation, the audit deadline and the checks that could not run; part of the paid plans, Pro and Team, when gating is on; a free, email-verified account's one free full document covers its report for 24 hours), expense_audit_portfolio (audit_expense_portfolio: a PDF or XLSX of a tenant's audit of up to 50 landlord expense statements at once: key figures, totals by landlord, the locations ranked by dollars at stake, dispute deadlines in date order, repeated issues and one short section per location; the XLSX has one row per location and one row per finding; part of the paid plans, Pro and Team, when gating is on, and the free account's one free full document does not cover it; it counts one report for each location audited in full, locations with invalid input are not counted and the answer's reports shows the month's use), estoppel_certificate (estoppel_prepare: a PDF tenant estoppel certificate, numbered statements each with its lease quote and page, the tenant's answer beside each when you send them as context.estoppel_answers and who has recorded reliance as context.estoppel_reliers, a signature block that says the tenant's signature, not FactLot, makes it an estoppel, and the document fingerprints; part of the paid plans, Pro and Team, when gating is on; a free, email-verified account's one free full document covers its report for 24 hours), rent_roll_certificate (rent_roll_certify: a PDF or XLSX rent roll certificate, each line with its grade, the roll and lease figures, quotes and pages, a differences section, the rollover schedule, WALT and the document fingerprints; send who has recorded reliance as context.rent_roll_reliers; the certifying party's certification, not FactLot, is what a lender or buyer relies on; part of the paid plans, Pro and Team, when gating is on; a free, email-verified account's one free full document covers its report for 24 hours), property_sheet (pdf or xlsx: one page for one parcel, arguments { urn } only, the parcel's FactLot ID from registry_search_address: FactLot runs registry_get on it as your own call, so it costs one lookup and is gated like any registry_get of yours (free: the current roll year only, with a line naming the earlier years a Pro or Team key adds), then lays out the address, IDs, owner of record when the registry holds an organisation (or why there is none), the linked registered entity as the registry's match with its confidence, values by roll year with each file's as-of date, building and land facts, for a Texas parcel the next general protest deadline under Tex. Tax Code §41.44(a)(1) labelled as the general rule, and every source with its licence or attribution; no context is accepted). Every kind but lease_abstract requires the tool's arguments: FactLot runs the tool on them and builds the document from that run only, so its numbers are FactLot's results; a result sent as input is not used (the answer's result_mismatch says whether it differs). lease_abstract takes doc_abstract_lease's arguments, or its result as input: built from a supplied result, every page says the abstract values are as supplied by the caller and FactLot did not recompute them. Caller text (title, prepared_for, citation locators, comparable ids, labels and reasons, tx_protest_deadline item ids) is printed as supplied by the caller; advice or valuation wording in it is refused (TextNotAllowed). One report per request (one report_create per MCP batch). Returns download_url (works 24 hours), expires_at, bytes, sha256, filename and billed (true when this request counted against the month; every refusal has billed: false); inline: true also returns files under 1 MiB as base64. Counts against reports_per_month (free: 3 a month per IP address; Pro 50; Team 1,000); input errors are free.

    Annotations the server declares

    • title: Create a report
    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: false
    • openWorldHint: false
  • rules_tx_property_tax_protest_deadlineDeadline to file a Texas property tax protest with the appraisal review board: the later of May 15 or 30 days after the notice of appraised value was delivered, rolled past weekend…No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Deadline to file a Texas property tax protest with the appraisal review board: the later of May 15 or 30 days after the notice of appraised value was delivered, rolled past weekends and Texas legal holidays, plus a conservative file-by date (the last business day on or before the unadjusted deadline). status says where as_of (default: today in US Central time; echoed in as_of) falls: not_yet_open before January 1 of the tax year, open from then through the deadline, passed after it; days_remaining counts the days to the deadline (null once passed), and once passed next_season gives the same rule for the next tax year (May 15 basis). notes says what the rule does not cover (late protests, other §41.44 deadlines, §25.25 corrections) and which sections to read. Returns the statute citations, the rule's verification status and a disclaimer. Information about a statutory deadline: not legal or tax advice and not a substitute for the advice of an attorney. Verification status: verified. The answer carries rule rules.tx.property_tax_protest_deadline@2 and the statutes it rests on.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Texas property tax protest deadline
  • rules_tx_property_tax_protest_deadline_batchTexas property tax protest deadlines for many notices in one call (1–200 items, e.g.No findings

    Why it is in this group

    • Nothing in its name or description says what it does, so it is counted as a write.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
    Description, as the server wrote it
    Texas property tax protest deadlines for many notices in one call (1–200 items, e.g. a consultant's client list): each item's id, tax year and, if known, notice delivery date. Each result is exactly what rules_tx_property_tax_protest_deadline returns for that item under rule rules.tx.property_tax_protest_deadline@2 (deadline, unadjusted deadline, governing basis, conservative file-by date, skipped days, status, days_remaining, next_season, the item's warnings), echoed with its id; an item with invalid input carries its error and the others are still answered. One as_of (default: today in US Central time) applies to every item. summary counts the items per deadline, earliest first, with the earliest and latest deadline, and per status (by_status). The as_of used, the notes, the statute citations, the rule's verification status and the disclaimer appear once (value.as_of, value.notes, value.rule, value.disclaimer). A rules tool: never a registry lookup. Information about statutory deadlines: not legal or tax advice and not a substitute for the advice of an attorney; it does not say whether to protest. Verification status: verified. The answer carries rule rules.tx.property_tax_protest_deadline_batch@1 and the statutes it rests on.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Texas property tax protest deadlines (batch)
Only reads 19Show 19 read-only tools
  • calc_compare_proposalsRank 2–20 lease proposals with different terms, sizes, free rent, TI, concessions, and operating-expense structures (gross, NNN, base year, expense stop, with opex growth) on one d…No findings

    Why it is in this group

    • The name "calc_compare_proposals" says "compare": a word for reading.
    Description, as the server wrote it
    Rank 2–20 lease proposals with different terms, sizes, free rent, TI, concessions, and operating-expense structures (gross, NNN, base year, expense stop, with opex growth) on one discounted $/SF/yr basis, per rentable or usable SF (load factor): effective occupancy cost (tenant perspective) or discounted net effective rent after landlord-borne opex and optional pre-commencement downtime (landlord perspective). Cash flows monthly in advance at annual rate ÷ 12; warns when the usable-SF ranking differs and when terms differ. schedule.unit (kw, mw, unit, acre, key, room, space, the same for every proposal) ranks per kW, MW, unit, acre, key, room, or space instead of per SF. Deterministic: the same inputs always give the same result, with formula ID calc.compare_proposals@2, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Compare lease proposals
  • calc_lease_vs_ownCompare leasing with buying over a fixed horizon: NPV of each case's annual cash costs (rent, free rent, tenant expenses, and move-in costs versus equity, debt service, owner expen…Reads private dataNo findings

    Why it is in this group

    • The description's first verb is "compare": a word for reading.

    What it can reach

    • private-dataThe description's first sentence says "compare" and "table": it reads databases.
    Description, as the server wrote it
    Compare leasing with buying over a fixed horizon: NPV of each case's annual cash costs (rent, free rent, tenant expenses, and move-in costs versus equity, debt service, owner expenses, reserves, and a sale at the horizon by stated value or exit cap rate), optionally after tax, with the year-by-year tables and the break-even residual value at which owning equals leasing. Deterministic: the same inputs always give the same result, with formula ID calc.lease_vs_own@2, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Lease vs. own
  • calc_multifamily_rent_rollAnalyze an apartment rent roll (up to 5,000 units with status occupied, notice, vacant, model, or down; market and lease rent per month; lease dates; concessions as weeks free, mon…No findings

    Why it is in this group

    • The description's first verb is "analyze": a word for reading.
    Description, as the server wrote it
    Analyze an apartment rent roll (up to 5,000 units with status occupied, notice, vacant, model, or down; market and lease rent per month; lease dates; concessions as weeks free, months free, or dollars): gross potential rent, vacancy, non-revenue units, loss to lease, concessions amortised straight-line over each lease term, effective rent income, physical and economic occupancy, effective rent per unit and per SF, a lease expiration schedule by month, and a per-unit-type summary. Single-lease mode (lease instead of units) returns the effective monthly rent after a weeks-free or months-free concession. Deterministic: the same inputs always give the same result, with formula ID calc.multifamily_rent_roll@1, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Multifamily rent roll
  • calc_renew_vs_relocateCompare renewing in place with moving over one horizon: NPV and discounted effective rent of each case (rent schedule with free rent, TI net of allowance, fees, moving, downtime, d…No findings

    Why it is in this group

    • The description's first verb is "compare": a word for reading.
    Description, as the server wrote it
    Compare renewing in place with moving over one horizon: NPV and discounted effective rent of each case (rent schedule with free rent, TI net of allowance, fees, moving, downtime, double rent, termination, and disruption costs), the starting rent at which they break even, and year-by-year cash flows. Deterministic: the same inputs always give the same result, with formula ID calc.renew_vs_relocate@2, every step, and any warnings.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Renew vs. relocate
  • doc_compareCompare values another document asserts (an estoppel, rent roll or offering memo, mapped by you to predicates) with verified lease facts from doc_abstract_lease or doc_verify_facts…Reads private dataNo findings

    Why it is in this group

    • The name "doc_compare" says "compare": a word for reading.

    What it can reach

    • private-dataThe name "doc_compare" says "compare" and "doc": it reads documents.
    • private-dataThe description's first sentence says "compare" and "document": it reads documents.
    Description, as the server wrote it
    Compare values another document asserts (an estoppel, rent roll or offering memo, mapped by you to predicates) with verified lease facts from doc_abstract_lease or doc_verify_facts. Each asserted value comes back match, mismatch (with the differing fields), not_in_lease, or ambiguous_in_lease (two current lease facts disagree), with the controlling lease citation (document, page, quote). Values compare exactly as JSON: numbers by value, strings exactly (a name written differently is a mismatch), and for an object only the fields you assert. Facts another fact supersedes are ignored. When the lease fact is an extracted fact whose quote does not confirm part of its value (a true/false value, an enum), the result lists those parts in lease_not_checked. The lease facts must pass fact_validate; FactLot does not re-read the lease, so pass the facts its document tools returned.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Compare asserted values with a lease
  • fact_package_statusRead the status of a package sent with fact_package_send: for each item, whether the counterparty agreed or disputed it (with their reason and who answered), which items the sender…Reads text others writeNo findings

    Why it is in this group

    • The name "fact_package_status" says "status": a word for reading.

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    Description, as the server wrote it
    Read the status of a package sent with fact_package_send: for each item, whether the counterparty agreed or disputed it (with their reason and who answered), which items the sender accepted, the reliances recorded, and the package's integrity. For an estoppel it also gives the statement number and text, the tenant's own version where it gave one, the signer's name and title with the electronic signature record (time in UTC, consent wording version and text, SHA-256 of the signed statements, and whether they are unchanged), reliers (each lender or buyer that recorded reliance, its role, the statements and when), and for each relying-party link the first and last time it was opened and how many times, with the name, firm and role given once reliance is recorded and the statements confirmed and disputed at that moment. Give the sender's private status_link (or package_id with token). Needs no account and costs nothing; it reads only your own package, and the counterparty's link cannot read it. The answer also gives attestation (attested or unattested, with the line the pages show: computed by FactLot and unchanged since, or not verified) and package_signature (the outcome of checking the signature fact_package_create's holder made, when there is one). The answer says whether the package was sent in test mode (test_mode true: it does not count in the FactLot record's measure) or for real.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: true
    • destructiveHint: false
    • title: Read a sent fact package's status and answers
  • fact_validateCheck one FactLot fact against the published fact schema (spec/schemas/fact.json): subject URN, dotted predicate, as_of date, provenance, optional rights; unknown fields are reject…No findings

    Why it is in this group

    • The name "fact_validate" says "validate": a word for reading.
    Description, as the server wrote it
    Check one FactLot fact against the published fact schema (spec/schemas/fact.json): subject URN, dotted predicate, as_of date, provenance, optional rights; unknown fields are rejected at every level. Rules: every fact except a computed one cites at least one source (an inferred fact cites the evidence it was inferred from); a computed fact names its formula or cites a source; valid_from is not after valid_to; a source span does not end before it starts. For Ingest predicates (lease.expiration_date, lease.rentable_area_sf, lease.base_rent_schedule, ...) the value must have the predicate's shape and every source quote must state the value, by the same written-form rules as doc_verify_facts (a quote saying "2,500 square feet" does not support 1000). Not checked: that the quote is in the document, that the document or sha256 exists, that supersedes IDs exist, and quotes for other predicates; doc_verify_facts with the document's pages checks the quote. Returns { valid, issues }. Meant for checking facts before they are stored or shared.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Validate a FactLot fact
  • fetchFetch the full text of a FactLot knowledge document by the id returned from search.Reads private dataNo findings

    Why it is in this group

    • The name "fetch" says "fetch": a word for reading.

    What it can reach

    • private-dataThe description's first sentence says "fetch" and "document": it reads documents.
    Description, as the server wrote it
    Fetch the full text of a FactLot knowledge document by the id returned from search. With a registry loaded, a registry fact's source document_id (fl:document:sha256-…) returns that source snapshot: publisher, dataset, URL, retrieval date, as-of date, licence, AI-use and redistribution rights, file name and size.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Fetch a FactLot knowledge document
  • id_parseValidate a FactLot URN and return its kind, components, and short ID.No findings

    Why it is in this group

    • The description's first verb is "validate": a word for reading.
    Description, as the server wrote it
    Validate a FactLot URN and return its kind, components, and short ID.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Parse a FactLot ID
  • lease_comp_validateCheck a lease comparable one firm shares with others, as a FactLot fact with predicate lease.comp, and derive its net effective rent.No findings

    Why it is in this group

    • The name "lease_comp_validate" says "validate": a word for reading.
    Description, as the server wrote it
    Check a lease comparable one firm shares with others, as a FactLot fact with predicate lease.comp, and derive its net effective rent. The fact goes through fact_validate (schema, citations, rights), then the comp's own rules: shape (spec/schemas/fact.json LeaseComp: area, commencement, term or expiration, starting or current rent, escalation, free rent, TI, leasing commissions, expense basis, confirming party, verification), an expiration after commencement, a term that matches the dates, free rent within the term, no email, phone or ID number in the text. A comp names no person or company. With the default share profile it also needs the deal's starting rent, escalations, free rent, TI, property type, an expense basis other than unknown, and stated rights. When valid and a starting rent is given, ner.fact is a computed lease.net_effective_rent fact (formula calc.net_effective_rent@1, straight-line and discounted at discount_rate_annual, on the comp's own expense basis; no conversion between full service and net). Returns { valid, issues, warnings, ner }. Meant for checking a comp before it is shared.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Validate a lease comp and derive its NER
  • registry_getEverything the FactLot Registry knows about a parcel, legal entity, or building: status, current cited facts (owner of record, situs address, property category, areas, values), own…No findings

    Why it is in this group

    • The name "registry_get" says "get": a word for reading.
    Description, as the server wrote it
    Everything the FactLot Registry knows about a parcel, legal entity, or building: status, current cited facts (owner of record, situs address, property category, areas, values), ownership links, external IDs, and for a parcel the building it belongs to (condominium units roll up into one building). value_history (parcels) lists market and appraised value by tax year from every snapshot, newest first, each value cited; a tax year two snapshots value differently appears once per value. as_of gives the facts as they stood on a date. A parcel's recorded sales and documents (transaction.sale, transaction.deed, debt.mortgage_recorded, debt.mortgage_satisfaction_recorded; one fact each, newest first) are kept to the newest 50 in facts and in history; when there are more, recorded_facts_truncated gives what was kept and the totals by predicate with the years left out, and registry_recorded_facts reads them all. warnings flags area facts that look wrong (building area above 40 times the land area, a land area of 0); the facts themselves are as the source publishes them. A building record can carry debt.cmbs_loan facts (one per CMBS trust holding a note of its loan: balance, rate, maturity, payment status, servicer, special-servicing dates, NOI, DSCR) and an entity record can carry entity.parent facts (the SEC registrant that lists it as a subsidiary in Exhibit 21), each cited to its SEC filing. A miss says whether the county or entity registry is loaded. On the hosted API, a call without a Pro or Team key gets the current roll year only in value_history and history, and of a parcel's recorded sales and documents only the newest sale or deed, with the count and years of the rest (the answer says so); a Pro or Team key gets every year and the newest 50 recorded facts. Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Get a registry record
  • registry_lookup_batchUp to 50 parcels in any loaded state in one call (a loan tape, a broker's list, a client's accounts), each item naming one parcel by exactly one of: urn (a FactLot parcel URN); apn…1 finding

    Why it is in this group

    • The name "registry_lookup_batch" says "lookup": a word for reading.

    Findings

    • Info

      The description is 4,413 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...ecord facts about commercial property, each cited to the source file and record)."

      What this check looks for
    Description, as the server wrote it
    Up to 50 parcels in any loaded state in one call (a loan tape, a broker's list, a client's accounts), each item naming one parcel by exactly one of: urn (a FactLot parcel URN); apn with state_fips and county_fips (3 digits), or with a 5-digit county_fips alone (as registry_resolve_parcel takes them); or address with the optional place fields registry_search_address takes (state, city, postal_code, county_fips). ref, your own row id, is echoed back. Each item is answered as the single tools answer it: registry_get for a URN, registry_resolve_parcel for an account, registry_search_address (20 results) for an address, then registry_get on the URN found. results lists one entry per item in input order (index, ref, by: urn, account or address, status): found gives parcel, a compact row: urn, label, record_status, account, property_category, use_code, improvement_area_sf, year_built, value (the newest roll year's entry of registry_get's value_history: tax_year, market and appraised value, as_of, value types, file and each value's citation), owner_of_record (the owner of record fact exactly as registry_get gives it, with the entity linked to it, or null; organisations only, natural persons are never stored), citations (each field's fact id, document id, locator and as-of date), sources (each cited document with its title, URI, read date, licence, ai_use and redistribution) and warnings; an address item also gives match (exact, and how many parcels the search returned). not_found gives message and coverage exactly as the single tool's miss gives them (the reason: county or state not loaded, street or house number not held); a URN not in the registry also carries registry_get's NotFound as error, with the same coverage registry_resolve_parcel gives for that county. An address is found when exactly one search result is exact, or when the search returned one parcel and it is in the place the address names; exact means the same street address and unit (the label's street line, normalized as the address index normalizes it, 'Street' = 'St', is the query's, a unit written as a bare number after the street counting as the unit, and both name the same unit or neither does) in the place the address names (when it names a city or ZIP code, in its text or the city and postal_code fields, the label has that ZIP code or city). Otherwise it is ambiguous, with reason: several_places (the address names no city or ZIP code and the results are in more than one county; a parcel is never picked across counties), same_address (several parcels have the address exactly: separate accounts at one address), place_differs (the one parcel found is not shown in the place named) or no_exact_match; candidates lists up to 5 of the search's result cards (the exact ones first; never an owner), with matches and a message. error gives the item's own error, so it does not fail the others: a wrong combination of fields (two of urn, apn and address, a field another mode takes, a missing county), a malformed or non-parcel URN, or an item the registry could not answer (RegistryUnavailable; after 5 such items the rest are not sent and say not attempted, and when no item at all is answered the whole call fails with RegistryUnavailable). An item whose fields fail the input schema (a wrong type, an unknown field, an empty ref, a URN over 500 characters) fails the whole call with InputError, uncharged. summary counts items, found, not_found, ambiguous and errors. On the hosted API this batch tool needs a Pro or Team key, and each item counts as the one call that identifies its parcel would count on its own (registry_resolve_parcel for an account or a URN, registry_search_address for an address): a found or ambiguous item counts one registry lookup; an item that finds nothing inside loaded coverage, a URN not in the registry included, is an answer that finds nothing (not counted up to an hourly limit, counted past it); a miss outside loaded coverage and an item with an error count none; the call needs one lookup per item left in the allowance to run, and one MCP request may make at most 500 registry service calls (an item counts 2), so a batch call past that is refused whole, uncharged. Facts as the sources publish them: not an appraisal, a valuation opinion or advice. Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Look up many parcels (batch)
  • registry_resolve_externalFind the FactLot URN for an ID from another system, e.g.No findings

    Why it is in this group

    • The name "registry_resolve_external" says "resolve": a word for reading.
    Description, as the server wrote it
    Find the FactLot URN for an ID from another system, e.g. system "sdat_account" (a Maryland SDAT account number), "hcad_account" (a Harris CAD account number), "nyc_bbl" (a New York City borough-block-lot), "la_ain" (a Los Angeles County Assessor ID number), "sf_parcel", "boston_pid", "denver_schednum" or "ct_cama_link". registry_coverage lists the systems a registry holds. Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Resolve an external ID
  • registry_resolve_parcelFind a parcel's FactLot URN from state FIPS, county FIPS, and APN, if the registry holds it.No findings

    Why it is in this group

    • The name "registry_resolve_parcel" says "resolve": a word for reading.
    Description, as the server wrote it
    Find a parcel's FactLot URN from state FIPS, county FIPS, and APN, if the registry holds it. A miss returns urn null with coverage: whether that county is loaded in this registry. Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Resolve a parcel in the registry
  • registry_search_addressFind commercial parcels by situs address.1 finding

    Why it is in this group

    • The name "registry_search_address" says "search": a word for reading.

    Findings

    • Info

      The description is 3,113 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...ecord facts about commercial property, each cited to the source file and record)."

      What this check looks for
    Description, as the server wrote it
    Find commercial parcels by situs address. Matches whole words of the street line (the part before the first comma), normalized: 'Street' = 'St', 'Northwest' = 'N.W.' = 'NW'; a suite, unit, '#' or floor is ignored. A numbered street matches however it is written ('Fifth Avenue' = '5th Ave' = '5 Avenue', 'Forty-Second St' = '42nd St'), and a house number matches a lot whose address is a range holding it ('350 5th Ave' finds '338-350 5 AVENUE'): a range whose ends are both odd or both even holds only the numbers of that parity, one side of the street, and one whose ends differ holds every number between. Give a house number ('1200 Main', '1200 Main St NW') or a street name with its type ('Main St'); a city or a part of a word alone matches nothing. Give the place after the first comma ('1200 Main St, Houston, TX 77002') or as state, city, postal_code and county_fips: a state named after a city (or given as state) keeps the answer to that state, and county_fips to that county, before the limit is applied; the ZIP code and the city only order the matches (same ZIP code first, then same city, then the rest) and never filter; a bare state with no city in the query, and suite, floor or box text, are ignored. Within each place group, matches of the street as written come first, then other spellings, then house-number ranges, an exact address first, then the parcel whose label names the unit the query names (a unit never filters), then by label; parcels sharing one label come largest improvement area first (a parcel the roll gives no improvement area after those it gives one), those the roll records as land only last, then in URN order. Each result carries what tells parcels at one address apart: the district's account or lot number, property category, use code, improvement area, land_only (true when the roll records an improvement area of 0, false when it records an improvement, null when it records neither), and the unit its label names; never an owner. When the query names a city or ZIP code, a result whose own label does not name it and whose county is outside every county that place lies in, in the result's state (from the US Census Bureau's city and ZIP code tables and the registry's own labels; a place the tables do not know there is not judged), carries place_differs: true, and when every result does, note says the matches are outside the counties the place lies in and place_counties lists them. No match returns coverage: what this registry holds (in the named state when there is one), and in reason and message why (a city or ZIP code that lies wholly in counties the registry does not hold is said first, with those counties and the ones held in that state; a city or ZIP code never filters, so that search still covered every county held): the state or county is not loaded, the street is not held, the house number is held but not with the query's other words, or the street is held but not that house number (with the nearest numbers held on it). Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Search parcels by address
  • registry_search_entitiesFind registered business entities (LLCs, LPs, corporations) by name.No findings

    Why it is in this group

    • The name "registry_search_entities" says "search": a word for reading.
    Description, as the server wrote it
    Find registered business entities (LLCs, LPs, corporations) by name. Natural persons are not searchable. Requires a FactLot Registry (public-record facts about commercial property, each cited to the source file and record).

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: Search legal entities by name
  • schema_getReturn the FactLot data model (Fact, Provenance, SourceRef, Rights, Relationship) as JSON Schema.No findings

    Why it is in this group

    • The name "schema_get" says "get": a word for reading.
    Description, as the server wrote it
    Return the FactLot data model (Fact, Provenance, SourceRef, Rights, Relationship) as JSON Schema.

    Annotations the server declares

    • readOnlyHint: true
    • idempotentHint: true
    • openWorldHint: false
    • destructiveHint: false
    • title: FactLot JSON Schemas

This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.

Exfil surface

Which tools read private data, which read text someone else can write, and which can send data out. One of each is a path a prompt injection can use.

6 paths in this server alone from private data, through text someone else wrote, to a way out. Denying 1 outbound tool breaks every one.

Reads private data3
Reads files, mail, messages, databases, secrets, calendars or repositories.
Reads text others write2
Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
Can send data out1
Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.
deny to break every path
{
  "permissions": {
    "deny": [
      "mcp__factlot__fact_package_send"
    ]
  }
}

Other servers in the same agent add their own tools to these paths. Audit your whole setup to see them together.

Remote posture

What anyone on the internet sees at the remote endpoint: an anonymous initialize and tools/list, and the public OAuth metadata.

Tools listed without sign-in

Looked at Nov 18, 2026, 16:22 UTC, unchanged since Nov 18, 2026. No credentials were sent and no tool was called. What is looked at.

Reachable
yes
HTTP status of an anonymous initialize
200
tools/list served without credentials
yes
Tools that change things, listed without credentials
43
Protocol version
2025-06-18
WWW-Authenticate on 401
not observed
resource_metadata in WWW-Authenticate
not observed
RFC 9728 protected resource metadata
https://api.factlot.com/.well-known/oauth-protected-resource/mcp
Authorization servers
https://api.factlot.com
Legacy HTTP+SSE transport
no
Dynamic client registration
yes
Client ID metadata documents
yes
  • WarnThe whole server

    The endpoint served tools/list without credentials, and the list has 43 tools that change things (43 write). Whether a call also works without credentials was not tested: tools are never called.

    Matched text

    https://api.factlot.com/mcp: tools/list returned 61 tools to an anonymous client, including calc_rent_schedule (write), calc_net_effective_rent (write), calc_percentage_rent (write), calc_cam_reconciliation (write), calc_cam_reconciliation_batch (write) and 38 more

    What this check looks for

Posture changes

  1. Nov 18, 2026
    • Tools that change things, listed without credentials: 42 to 43
  2. Nov 11, 2026
    • Tools that change things, listed without credentials: 16 to 42
    • RFC 9728 protected resource metadata: none to https://api.factlot.com/.well-known/oauth-protected-resource/mcp
    • Authorization servers: none to https://api.factlot.com
    • Dynamic client registration: not observed to yes
    • Client ID metadata documents: not observed to yes

Findings

Matches of fixed checks, with the text they matched.

3 warn11 info
  • The server marks "calc_rent_schedule" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

    Matched text

    annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The name "calc_rent_schedule" says "schedule": a word for a change.

    What this check looks for
  • The server marks "calc_amortization_schedule" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

    Matched text

    annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The name "calc_amortization_schedule" says "schedule": a word for a change.

    What this check looks for
  • The server marks "fact_package_create" read-only (readOnlyHint: true), and its own name, description or input give the class "write".

    Matched text

    annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":false}; The name "fact_package_create" says "create": a word for a change.

    What this check looks for
  • The description is 2,248 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...esult, with formula ID calc.tx_unequal_appraisal@1, every step, and any warnings."

    What this check looks for
  • The description is 2,623 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...ded for 24 hours; Pro and Team include every document with the monthly allowance."

    What this check looks for
  • The description is 2,210 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...(report_units); this tool call is not counted, as audit_expense_statement is not."

    What this check looks for
Show 8 more findings
  • The description is 2,617 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...ded for 24 hours; Pro and Team include every document with the monthly allowance."

    What this check looks for
  • The description is 2,452 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...ded for 24 hours; Pro and Team include every document with the monthly allowance."

    What this check looks for
  • The description is 3,722 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...tatus and its line). fact_package_status reads the answers and the reliers later."

    What this check looks for
  • The description is 3,113 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...ecord facts about commercial property, each cited to the source file and record)."

    What this check looks for
  • The description is 3,917 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...ecord facts about commercial property, each cited to the source file and record)."

    What this check looks for
  • The description is 2,150 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...ecord facts about commercial property, each cited to the source file and record)."

    What this check looks for
  • The description is 4,413 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...ecord facts about commercial property, each cited to the source file and record)."

    What this check looks for
  • The description is 5,561 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...onth (free: 3 a month per IP address; Pro 50; Team 1,000); input errors are free."

    What this check looks for

History

Changes to its tool definitions and to its remote endpoint, newest first.

  1. Tools9 added, 43 changed.+9 added~43 changedView diff
  2. PostureRemote posture changed. Tools that change things, listed without credentials: 42 -> 43.View diff
    • Tools that change things, listed without credentials: 42 to 43
  3. PostureRemote posture changed. Tools that change things, listed without credentials: 16 -> 42; RFC 9728 protected resource metadata: none -> https://api.factlot.com/.well-known/oauth-protected-resource/mcp; Authorization servers: none -> https://a…View diff
    • Tools that change things, listed without credentials: 16 to 42
    • RFC 9728 protected resource metadata: none to https://api.factlot.com/.well-known/oauth-protected-resource/mcp
    • Authorization servers: none to https://api.factlot.com
    • Dynamic client registration: not observed to yes
    • Client ID metadata documents: not observed to yes
  4. Tools10 added, 7 changed.+10 added~7 changedView diff
  5. Tools6 added, 7 changed.+6 added~7 changedView diff
  6. Tools13 added, 11 changed.+13 added~11 changedView diff
  7. Tools5 added, 9 changed.+5 added~9 changedView diff

Install

Pinned to version 0.5.0. Values in angle brackets are yours to fill in.

claude mcp add --transport http factlot https://api.factlot.com/mcp
claude mcp add --transport http factlot https://api.factlot.com/mcp

Run in a terminal.

Claude Code permissions

Allow the reads, ask before a change or a way out, deny the deletes. Paste into .claude/settings.json.

.claude/settings.json
{
  "permissions": {
    "allow": [
      "mcp__factlot__calc_compare_proposals",
      "mcp__factlot__calc_multifamily_rent_roll",
      "mcp__factlot__calc_lease_vs_own",
      "mcp__factlot__calc_renew_vs_relocate",
      "mcp__factlot__id_parse",
      "mcp__factlot__fact_validate",
      "mcp__factlot__doc_compare",
      "mcp__factlot__fact_package_status",
      "mcp__factlot__lease_comp_validate",
      "mcp__factlot__lease_comp_search",
      "mcp__factlot__registry_get",
      "mcp__factlot__registry_resolve_parcel",
      "mcp__factlot__registry_resolve_external",
      "mcp__factlot__registry_search_address",
      "mcp__factlot__registry_search_entities",
      "mcp__factlot__registry_lookup_batch",
      "mcp__factlot__search",
      "mcp__factlot__fetch",
      "mcp__factlot__schema_get"
    ],
    "ask": [
      "mcp__factlot__calc_rent_schedule",
      "mcp__factlot__calc_net_effective_rent",
      "mcp__factlot__calc_percentage_rent",
      "mcp__factlot__calc_cam_reconciliation",
      "mcp__factlot__calc_cam_reconciliation_batch",
      "mcp__factlot__calc_capex_recovery",
      "mcp__factlot__calc_debt_sizing",
      "mcp__factlot__calc_lease_remeasurement_asc842",
      "mcp__factlot__calc_prepayment_and_defeasance",
      "mcp__factlot__calc_walt",
      "mcp__factlot__calc_npv",
      "mcp__factlot__calc_irr",
      "mcp__factlot__calc_hotel_metrics",
      "mcp__factlot__calc_dcf_pro_forma",
      "mcp__factlot__calc_direct_cap",
      "mcp__factlot__calc_amortization_schedule",
      "mcp__factlot__calc_equity_waterfall",
      "mcp__factlot__calc_exchange_1031",
      "mcp__factlot__calc_critical_dates",
      "mcp__factlot__calc_reconcile_dcf",
      "mcp__factlot__calc_reconcile_cam",
      "mcp__factlot__calc_tx_unequal_appraisal",
      "mcp__factlot__rules_tx_property_tax_protest_deadline",
      "mcp__factlot__rules_tx_property_tax_protest_deadline_batch",
      "mcp__factlot__id_parcel",
      "mcp__factlot__id_entity",
      "mcp__factlot__id_space",
      "mcp__factlot__id_building",
      "mcp__factlot__id_document",
      "mcp__factlot__doc_verify_facts",
      "mcp__factlot__doc_abstract_lease",
      "mcp__factlot__doc_verify_claim",
      "mcp__factlot__doc_spread_statement",
      "mcp__factlot__doc_check_use_conflict",
      "mcp__factlot__doc_predicates",
      "mcp__factlot__fact_package_create",
      "mcp__factlot__fact_package_verify",
      "mcp__factlot__audit_expense_statement",
      "mcp__factlot__audit_expense_portfolio",
      "mcp__factlot__estoppel_prepare",
      "mcp__factlot__rent_roll_certify",
      "mcp__factlot__fact_package_send",
      "mcp__factlot__registry_owned_by",
      "mcp__factlot__registry_recorded_facts",
      "mcp__factlot__registry_coverage",
      "mcp__factlot__registry_tx_equity_candidates",
      "mcp__factlot__registry_tx_equity_candidates_batch",
      "mcp__factlot__report_create"
    ],
    "deny": []
  }
}

Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. A read that can send data out (it takes a URL, say) asks first too. It assumes the server is installed as factlot. Review it before you use it.

Capability tags behind 5 rules
  • mcp__factlot__calc_lease_vs_own private-data
  • mcp__factlot__doc_compare private-data
  • mcp__factlot__fact_package_send untrusted-input, outbound
  • mcp__factlot__fact_package_status untrusted-input
  • mcp__factlot__fetch private-data

Pin

Keep the decision: fail CI when the tools change, follow every change, or show a badge.

Pin in CI

Fail a build when the tools you reviewed are not the tools the server serves today.

curl -s "https://radar.anlyon.com/api/v1/servers/io.github.9mp96ygnbk-alt/factlot/verify?hash=39296e1183c1bfaa03ac15cd3670e6ac30d52b9c65ce6662460e4d1f110a5551" | jq -e '.changed == false'

Pinned hash 39296e1183c1, from Nov 21, 2026, 15:18 UTC.

Watch

Get told when its tools change: a new tool, a new description, a new class.

https://radar.anlyon.com/servers/io.github.9mp96ygnbk-alt/factlot/changes.xml

An Atom feed. Works in any feed reader or Slack's RSS app.

Badge

Anlyon Radar badge for io.github.9mp96ygnbk-alt/factlot
[![Anlyon Radar](https://radar.anlyon.com/badge/io.github.9mp96ygnbk-alt/factlot.svg)](https://radar.anlyon.com/servers/io.github.9mp96ygnbk-alt/factlot)

Source: MCP Registry · repository. Synced Nov 7, 2026, 11:17 UTC.

FactLot MCP server: questions and answers

Answered from the scan above, not from the publisher's description.

Is the FactLot MCP server safe?

No scan can promise that. What its definitions show: 67 tools, 48 of which can change, send or delete something, and 14 findings from fixed checks. Allow the reads, ask before each change, and deny the deletes; the permissions block on this page does that.

Can the FactLot MCP server delete data?

No tool is classed as destructive: going by their definitions, none of them deletes, cancels or revokes anything.

Can the FactLot MCP server do things that cannot be undone?

Yes. 1 tool can send, pay, publish or merge: fact_package_send.

What tools does the FactLot MCP server have?

67 tools: 1 irreversible, 47 write, 19 read. fact_package_send, calc_rent_schedule, calc_net_effective_rent, calc_percentage_rent, calc_cam_reconciliation, calc_cam_reconciliation_batch, calc_capex_recovery, calc_debt_sizing and 59 more.

Does the FactLot MCP server need sign-in or OAuth?

Its remote endpoint listed its tools to an anonymous client, without sign-in.

Does the FactLot MCP server send data to outside services?

1 tool can carry data out (it sends, posts or publishes, or takes a recipient, channel or URL): fact_package_send.

Has the FactLot MCP server changed its tools?

Yes, 5 times since Anlyon Radar started tracking it, most recently on Nov 21, 2026. Every change is on this page and in its Atom feed.