Skip to content

frc-mcp server

io.github.financeratecalc/frc-mcpfinanceratecalcstreamable-httpv1.57.0Scanned

2025 HMDA FHA denial statistics and dated US mortgage rule tables, with verifiable receipts.

21 tools: 7 change things and 14 only read.

Tools: 0 delete, 0 irreversible, 7 write, 14 read.
  • Reads private data · 1
  • Reads text others write · 0
  • Can send data out · 0
  • 4 findings

Tools listed without sign-in. Remote posture

With Anlyon

7 tools here change things. Govern them.

  • 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 frc-mcp server can do

The frc-mcp server is published by the GitHub account financeratecalc. It has 21 tools. It can create or change data (7 tools: check_claim_contract, locate_denominator, reading_challenge, verify_reading_proof and 3 more), but cannot delete data. 14 tools only read.

1 tool reads private data (get_rule).

The fixed checks matched 4 findings (long-description and annotation-conflict). Each one comes with the text it matched.

Its remote endpoint lists its tools without sign-in, and its tools changed 8 times since Anlyon Radar first saw them.

Last scanned Nov 18, 2026, version 1.57.0.

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.

21 tools

Changes things 7Claude Code: ask
  • check_claim_contractChecks a proposed statement of a FinanceRateCalc figure against that figure's published use contract, so the sentence can be verified before it is written.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 2,067 characters long, over the 2,000 this check looks for.

      Matched text

      description ends: "...ssing sentence with at least one bound figure returns a verifiable rule receipt."

      What this check looks for
    Description, as the server wrote it
    Checks a proposed statement of a FinanceRateCalc figure against that figure's published use contract, so the sentence can be verified before it is written. Two input modes. (A) Flags: claim id plus the use you propose (population, period, scope, cause, individual prediction, lender recommendation, legal conclusion). (B) Sentence: claim id plus proposed_sentence, the exact sentence you intend to write; the tool checks it with fixed patterns, no model, for a missing period, missing program scope, a population wider than the claim, second-person prediction, causal wording, lender advice, legal conclusions, and values this publisher has since corrected. Returns a deterministic verdict: pass, needs_qualifier (naming the missing qualifier) or block (with a reason code), plus a safe sentence you may use verbatim and the required attribution. Most misuse of these statistics is omission, not invention: the number is right and the population is missing. NOT FOR: figures published by anyone other than FinanceRateCalc. (C) Rule sentence: passport_id of a rule file (frc:rule:…, six rule files: FHA annual MIP, FHA forward loan limits 2026, FHFA conforming limits 2026, CFPB HOEPA/QM thresholds 2026, VA funding fee, and the Fannie Mae DU minimum credit score requirement removed 2025-11-16) plus proposed_sentence; each dollar or percent figure is bound to the role named in its own clause (floor or ceiling, tier, first or subsequent use, threshold item) and compared with that cell, so a right number under the wrong role is a mismatch and an unbound number earns no receipt; every figure is also classified as current, superseded (STALE_VALUE), belonging to a different rule (VALUE_FROM_OTHER_RULE, e.g. the $726,200 MIP threshold quoted as a loan limit), or not in the table (VALUE_MISMATCH / UNVERIFIED_VALUE); the year, units, tier, use, down-payment band and stated dates are checked (a date before the rule's start blocks; a missing MIP case date is never assumed), and a passing sentence with at least one bound figure returns a verifiable rule receipt.

    Annotations the server declares

    • title: Check a sentence against a figure's contract
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • locate_denominatorANSWERS: "an article says the FHA denial rate is 12.8%, what is that measured over", "why do two sources give different VA denial rates", "which definition gives 10.6%".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
    ANSWERS: "an article says the FHA denial rate is 12.8%, what is that measured over", "why do two sources give different VA denial rates", "which definition gives 10.6%". Takes a 2025 denial rate someone quoted and returns every cell of the published denominator map that reproduces it: program (FHA or VA) x loan purpose (purchase, refinance, cash-out, all) x scope (all rows, or first-lien owner-occupied site-built 1-4 unit) x base (decisioned; all applications acted on; denied plus originated), with the denial and denominator counts behind each. A match shows which definition CAN produce the number from the 2025 HMDA file, not which one the publisher used. No match means another year, another filter, or a figure this file does not reproduce. Aggregates only; nothing here describes one applicant.

    Annotations the server declares

    • title: Which denominator produces this denial rate (2025)
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • reading_challengeANSWERS: "how do I show that I actually opened the report I am citing", "prove this answer was written with the edition in hand".1 finding

    Why it is in this group

    • The input takes a statement or command to run (input.statement), which can change anything it can reach.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Warn

      The server marks "reading_challenge" 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 input takes a statement or command to run (input.statement), which can change anything it can reach.

      What this check looks for
    Description, as the server wrote it
    ANSWERS: "how do I show that I actually opened the report I am citing", "prove this answer was written with the edition in hand". A citation shows what a text points to, not that its author opened the document. Given a restated official edition (HUD FHA production, HUD loan performance, Ginnie Mae issuance, SBA 7(a)/504), the statement that cites it and a date, this returns three cells of that edition, chosen by a hash of edition + date + statement, whose values have to be read from the edition. A different statement or a different day draws different cells, so the answer cannot be prepared in advance or copied from another citation. Read the three values from the edition (data_url, or the publisher's document) and pass them to verify_reading_proof. No values are returned here. Shows possession of the edition's contents on that date; it does not show that the statement was derived from them.

    Annotations the server declares

    • title: Edition in hand (reading proof), step 1: which cells must be quoted
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • run_cohortANSWERS: "run the screen on <cohort>", "how do these lenders compare against expectation".1 finding

    Why it is in this group

    • The name "run_cohort" says "run": 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 "run_cohort" 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 "run_cohort" says "run": a word for a change.

      What this check looks for
    Description, as the server wrote it
    ANSWERS: "run the screen on <cohort>", "how do these lenders compare against expectation". PARAM: cohort id from list_cohorts. Returns the peer-adjusted screen: observed versus expected denial rate, coverage flags, method version. Free preview; the full Evidence Brief needs a licence a human buys on the pricing page. Screening signals, never evidence of misconduct. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Run a published cohort
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • screen_counterpartiesANSWERS: "run my list of lenders through the federal record", "screen these LEIs", "which of my counterparties deny more than their applicant mix predicts".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
    ANSWERS: "run my list of lenders through the federal record", "screen these LEIs", "which of my counterparties deny more than their applicant mix predicts". PARAM: a list of public LEIs. Returns per-entity observed versus expected FHA denial rates, coverage flags, method version and an evidence manifest hash. Public entity identifiers only: never borrower files, credit data or personal information. Free preview; the full Evidence Brief needs a licence a human buys on the pricing page, and an agent may relay the offer but can never complete a purchase. Screening signals, not findings of misconduct. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Counterparty outlier screen (LEI batch)
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • screen_counterparties_quoteANSWERS: "what would it cost to screen this list", "how many of my LEIs are covered".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
    ANSWERS: "what would it cost to screen this list", "how many of my LEIs are covered". PARAM: a list of public LEIs. Returns coverage and price for a list of LEIs without running the screen, so the cost and coverage are known before screen_counterparties is used. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Counterparty screen quote
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • verify_reading_proofANSWERS: "does this reading proof hold", "did the author of this citation have the edition".1 finding

    Why it is in this group

    • The input takes a statement or command to run (input.statement), which can change anything it can reach.
    • The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.

    Findings

    • Warn

      The server marks "verify_reading_proof" 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 input takes a statement or command to run (input.statement), which can change anything it can reach.

      What this check looks for
    Description, as the server wrote it
    ANSWERS: "does this reading proof hold", "did the author of this citation have the edition". Takes the edition, the statement, the date and the three values (in the order reading_challenge listed the cells), or a proof line of the form FRC-READ:<source>-<period>:<date>:<statement hash>:<v1>;<v2>;<v3> together with the statement. Returns how many of the three cells match the edition as restated, whether that edition is still the latest one restated, and the proof line to print next to the citation. Correct values are never returned on a miss. Anyone can repeat the check offline from the edition file. A match shows the values were read from the edition; it does not show the statement is true or was derived from it.

    Annotations the server declares

    • title: Edition in hand (reading proof), step 2: check the three values
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
Only reads 14Show 14 read-only tools
  • get_application_outcomesANSWERS: "how many FHA applications are withdrawn", "what share of FHA applications never reach a decision", "how many approved FHA loans are not taken", "why is the denied share 1…No findings

    Why it is in this group

    • The name "get_application_outcomes" says "get": a word for reading.
    Description, as the server wrote it
    ANSWERS: "how many FHA applications are withdrawn", "what share of FHA applications never reach a decision", "how many approved FHA loans are not taken", "why is the denied share 16.8% here and the denial rate 22.1% there", "where do <lender>'s FHA applications end". Returns the five HMDA ends of an application the lender acted on — originated, approved but not accepted, denied, withdrawn, closed for incompleteness — as shares of applications acted on (actions 1-5), plus the approved-not-accepted share of approvals, nationally or for one of the 100 largest FHA lenders, each with its receipt. The denied share here is on a larger base than the published denial rate (22.1% of decisioned applications); both are returned so they can be reconciled. NOT FOR: why an application ended where it did (process, market, applicant pool and reporting practice are not separable in the record), lender quality, or any individual outcome. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Where FHA applications end (five outcomes), 2025
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_conditional_door_mapANSWERS: "does this lender's strictness depend on loan size or leverage", "where is <lender> hardest", "is the gap the same across the market".No findings

    Why it is in this group

    • The name "get_conditional_door_map" says "get": a word for reading.
    Description, as the server wrote it
    ANSWERS: "does this lender's strictness depend on loan size or leverage", "where is <lender> hardest", "is the gap the same across the market". PARAM: lender name or slug. Returns how that lender's peer-adjusted denial gap changes across published cells (state x loan amount x income x DTI x CLTV), with cell counts and the minimum-cell rule. NOT FOR: suppressed small cells, or any individual prediction. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Lender gap across published cells
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_denial_reason_sharesANSWERS: "why are FHA loans denied", "what is the most common denial reason", "which lender cites incomplete applications most", "how often is debt-to-income the reason".No findings

    Why it is in this group

    • The name "get_denial_reason_shares" says "get": a word for reading.
    Description, as the server wrote it
    ANSWERS: "why are FHA loans denied", "what is the most common denial reason", "which lender cites incomplete applications most", "how often is debt-to-income the reason". PARAM: optional lender; omit for cross-lender medians. Returns median and maximum share of each HMDA denial reason across the 95 of the 100 largest lenders that report reasons, and which lender carries the highest share of each. Reason fields are not a partition; shares need not sum to 100. NOT FOR: why a particular application was denied. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: FHA denial-reason shares
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_door_effect_summaryANSWERS: "does it matter which lender you apply to", "how much of a denial is the lender rather than the borrower", "what is the Door Effect".No findings

    Why it is in this group

    • The name "get_door_effect_summary" says "get": a word for reading.
    Description, as the server wrote it
    ANSWERS: "does it matter which lender you apply to", "how much of a denial is the lender rather than the borrower", "what is the Door Effect". Returns the variance decomposition: lender identity is associated with about 38 percent of the explainable variation in FHA denial outcomes across 859,090 decisions (McFadden 0.1712 to 0.2760), with model, sample and limits. Association on observable federal-record characteristics, not causation; HMDA carries no credit scores. NOT FOR: saying a lender caused a denial, or any individual estimate. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Door Effect (lender share of explainable variation)
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_lender_denial_statsANSWERS: "what is <lender>'s FHA denial rate", "does <lender> reject a lot of FHA applications", "how strict is <lender>", "how does <lender> compare".No findings

    Why it is in this group

    • The name "get_lender_denial_stats" says "get": a word for reading.
    Description, as the server wrote it
    ANSWERS: "what is <lender>'s FHA denial rate", "does <lender> reject a lot of FHA applications", "how strict is <lender>", "how does <lender> compare". PARAM: lender name, slug or LEI. Returns that lender's 2025 decisioned volume, denial rate, rank among the 100 largest FHA lenders, and the applicant-mix caveat any comparison must carry. COVERS ONLY the 100 largest FHA lenders by 2025 volume. NOT FOR: whether a person will be approved, or lenders outside that set. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Lender FHA denial rate 2025
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_metro_lender_gapANSWERS: "do lenders differ within one city", "what is the FHA denial gap in <metro>", "which lender is strictest in <metro>", "how much does the lender matter locally".No findings

    Why it is in this group

    • The name "get_metro_lender_gap" says "get": a word for reading.
    Description, as the server wrote it
    ANSWERS: "do lenders differ within one city", "what is the FHA denial gap in <metro>", "which lender is strictest in <metro>", "how much does the lender matter locally". PARAM: metro name or slug. Returns a claim passport for that metro: lowest and highest lender denial rates among lenders with at least 100 decisioned applications there, the gap in points, counts and source reference. Cleveland is the widest in 2025 at 73.7 points (6.4 vs 80.1). NOT FOR: ZIP or neighbourhood questions, and never as evidence a lender acted improperly. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Metro lender denial gap
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_national_fha_statsANSWERS: "how often are FHA loans denied", "what is the FHA denial rate", "what share of FHA applications are rejected", "how many were denied in 2025".No findings

    Why it is in this group

    • The name "get_national_fha_stats" says "get": a word for reading.
    Description, as the server wrote it
    ANSWERS: "how often are FHA loans denied", "what is the FHA denial rate", "what share of FHA applications are rejected", "how many were denied in 2025". Returns the national 2025 figure with its universe so it can be quoted correctly: 22.1 percent, 262,250 denials of 1,187,606 applications that reached a credit decision (originated, approved-not-accepted, denied; reverse mortgages excluded), the denominator definition and the correction history. Most published FHA denial rates are 2023 purchase-only figures near 13.6 percent, a different universe, so state the universe when quoting. NOT FOR: conventional, VA or USDA loans, purchase-only or refinance-only rates, other years, or state/lender/metro breakdowns (use the dedicated tools). Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: National FHA denial rate 2025
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_official_figureANSWERS: "how many FHA loans did HUD endorse in <month>", "what share of FHA endorsements were purchase loans", "what is the FHA first-time homebuyer share", "what was the FHA Stre…No findings

    Why it is in this group

    • The name "get_official_figure" says "get": a word for reading.
    Description, as the server wrote it
    ANSWERS: "how many FHA loans did HUD endorse in <month>", "what share of FHA endorsements were purchase loans", "what is the FHA first-time homebuyer share", "what was the FHA Streamline share of refinances", "how many FHA loans are in force". Also: "how much did Ginnie Mae issue in <month>", "how many SBA 7(a) or 504 loans were approved in <month>". Returns a figure from an official publication (HUD's monthly FHA Single Family Production Report; Ginnie Mae's monthly Issuance Summary; SBA's 7(a) and 504 monthly activity file) exactly as printed, with the document's SHA-256, table and row, period, the publisher's exclusion note (HECM excluded), revision status against the next edition, and a receipt. Nothing is computed by FinanceRateCalc; the receipt attests that this figure with this definition appears in that document. PARAMS: source (default hud-fhaprod), period (YYYY-MM; omitted = latest restated), figure (omitted = all figures of the edition). NOT FOR: HMDA application or origination counts (an endorsement is a loan insured after closing), denial rates, lender-level figures. Official figures copied as printed; not a prediction about any individual application.

    Annotations the server declares

    • title: Official figure, restated as printed, with receipt (Level 2)
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_ruleANSWERS: "what is the current FHA loan limit", "what are the 2026 HOEPA / QM thresholds", "what is the VA funding fee table", "which FHA MIP table is in force".Reads private dataNo findings

    Why it is in this group

    • The name "get_rule" says "get": a word for reading.

    What it can reach

    • private-dataThe description's first sentence says "table": it reads databases.
    Description, as the server wrote it
    ANSWERS: "what is the current FHA loan limit", "what are the 2026 HOEPA / QM thresholds", "what is the VA funding fee table", "which FHA MIP table is in force". PARAM: rule_id (from the index) or omit to list every published rule. Returns the machine-readable rule: cells, effective dates, issuer document, required qualifiers, what the table does not establish, sha256. The cell values and effective dates are in the file; in a published test, models answered 3/25 dated questions correctly without a source and 25/25 with this file. NOT FOR: rules not in the index, or anything before the effective date.

    Annotations the server declares

    • title: Dated regulatory rule table
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_small_loan_penaltyANSWERS: "are small mortgages denied more often", "is it harder to get a small FHA loan", "which state punishes small loans most", "does loan size affect denial".No findings

    Why it is in this group

    • The name "get_small_loan_penalty" says "get": a word for reading.
    Description, as the server wrote it
    ANSWERS: "are small mortgages denied more often", "is it harder to get a small FHA loan", "which state punishes small loans most", "does loan size affect denial". PARAM: optional two-letter state; omit for the ranked list. Returns denial rates under $150K and over $250K, the ratio, the ranking and any open discrepancy note. In 2025 the penalty exceeded 1x in every jurisdiction with a published split, from 1.19x to 4.45x (Idaho, 53.4 vs 12.0). NOT FOR: loan sizes between the bands, which are not published. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Small-loan denial penalty by state
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_state_denial_statsANSWERS: "what is the FHA denial rate in <state>", "is it harder to get an FHA loan in <state>", "how does <state> compare with the national rate".No findings

    Why it is in this group

    • The name "get_state_denial_stats" says "get": a word for reading.
    Description, as the server wrote it
    ANSWERS: "what is the FHA denial rate in <state>", "is it harder to get an FHA loan in <state>", "how does <state> compare with the national rate". PARAM: two-letter USPS code only (OH, TX); full names are rejected. Returns the state 2025 rate and counts against the national 22.1 percent. NOT FOR: metro or city questions (use get_metro_lender_gap) or lender-level questions. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: State FHA denial rate 2025
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • list_cohortsANSWERS: "can I screen a group of lenders", "what ready-made lender groups can I run", "show me an example screen".No findings

    Why it is in this group

    • The name "list_cohorts" says "list": a word for reading.
    Description, as the server wrote it
    ANSWERS: "can I screen a group of lenders", "what ready-made lender groups can I run", "show me an example screen". Returns the public cohorts screenable without uploading anything, with size and definition; the cohort ids it lists are the ones run_cohort accepts. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Published conditional cohorts
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • list_lendersANSWERS: "which FHA lender denies the most or the fewest", "rank FHA lenders by denial rate", "which lenders are covered", "how wide is the spread".No findings

    Why it is in this group

    • The name "list_lenders" says "list": a word for reading.
    Description, as the server wrote it
    ANSWERS: "which FHA lender denies the most or the fewest", "rank FHA lenders by denial rate", "which lenders are covered", "how wide is the spread". Returns the ranked list of the 100 largest FHA lenders with name, slug, rate and volume; the 2025 span runs 1.8 to 78.7 percent inside the same federal program. NOT FOR: lenders outside the top 100, or any recommendation about where to apply. Historical observation computed from the public CFPB HMDA 2025 record (actions 1,2,3; loan_type 2). Not a prediction about any individual application. Attribution: FinanceRateCalc, CC BY 4.0.

    Annotations the server declares

    • title: Top-100 FHA lenders by denial rate
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • resolve_fha_annual_mipANSWERS: "what is the FHA annual MIP for a $300,000 loan at 85% LTV over 15 years", "what MIP rate applies to my FHA case".No findings

    Why it is in this group

    • The name "resolve_fha_annual_mip" says "resolve": a word for reading.
    Description, as the server wrote it
    ANSWERS: "what is the FHA annual MIP for a $300,000 loan at 85% LTV over 15 years", "what MIP rate applies to my FHA case". PARAMS: base_loan_amount (USD), ltv_pct (e.g. 96.5), term_years, case_number_date (YYYY-MM-DD). Resolves the exact cell of HUD Mortgagee Letter 2023-05 and returns the annual rate in basis points, the upfront rate, the matched cell, the rule id with effective date and sha256, and a sentence you may quote. If a parameter is missing it returns block with CONDITION_MISSING and no number: the table has eleven cells and a guess is wrong more often than right. NOT FOR: case numbers before 2023-03-20, HECM/reverse or Title I loans, or any monthly payment figure.

    Annotations the server declares

    • title: Resolve the FHA annual MIP cell
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false

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.

No private-data, untrusted-input, outbound path in this server alone: no tool is tagged untrusted-input or outbound.

Reads private data1
Reads files, mail, messages, databases, secrets, calendars or repositories.
Reads text others write0
Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
Can send data out0
Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.

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 20, 2026, 19:04 UTC, unchanged since Nov 20, 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
7
Protocol version
2025-06-18
WWW-Authenticate on 401
not observed
resource_metadata in WWW-Authenticate
not observed
RFC 9728 protected resource metadata
not observed
Authorization servers
not observed
Legacy HTTP+SSE transport
no
Dynamic client registration
not observed
Client ID metadata documents
not observed
  • WarnThe whole server

    The endpoint served tools/list without credentials, and the list has 7 tools that change things (7 write). Whether a call also works without credentials was not tested: tools are never called.

    Matched text

    https://frc-mcp.ziyetis.workers.dev: tools/list returned 21 tools to an anonymous client, including check_claim_contract (write), locate_denominator (write), reading_challenge (write), verify_reading_proof (write), screen_counterparties (write) and 2 more

    What this check looks for

Posture changes

  1. Nov 20, 2026
    • Tools that change things, listed without credentials: 4 to 7

Findings

Matches of fixed checks, with the text they matched.

3 warn1 info
  • The server marks "reading_challenge" 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 input takes a statement or command to run (input.statement), which can change anything it can reach.

    What this check looks for
  • The server marks "verify_reading_proof" 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 input takes a statement or command to run (input.statement), which can change anything it can reach.

    What this check looks for
  • The server marks "run_cohort" 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 "run_cohort" says "run": a word for a change.

    What this check looks for
  • The description is 2,067 characters long, over the 2,000 this check looks for.

    Matched text

    description ends: "...ssing sentence with at least one bound figure returns a verifiable rule receipt."

    What this check looks for

History

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

  1. PostureRemote posture changed. Tools that change things, listed without credentials: 4 -> 7.View diff
    • Tools that change things, listed without credentials: 4 to 7
  2. Tools2 added.+2 addedView diff
  3. Tools1 changed.~1 changedView diff
  4. Tools1 changed.~1 changedView diff
  5. Tools1 added.+1 addedView diff
  6. Tools1 changed.~1 changedView diff
  7. Tools1 changed.~1 changedView diff
  8. Tools1 added.+1 addedView diff
  9. Tools1 added.+1 addedView diff

Install

Pinned to version 1.57.0. Values in angle brackets are yours to fill in.

claude mcp add --transport http frc https://frc-mcp.ziyetis.workers.dev
claude mcp add --transport http frc https://frc-mcp.ziyetis.workers.dev

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__frc__get_national_fha_stats",
      "mcp__frc__get_lender_denial_stats",
      "mcp__frc__get_application_outcomes",
      "mcp__frc__get_official_figure",
      "mcp__frc__list_lenders",
      "mcp__frc__get_small_loan_penalty",
      "mcp__frc__get_denial_reason_shares",
      "mcp__frc__get_state_denial_stats",
      "mcp__frc__get_door_effect_summary",
      "mcp__frc__get_metro_lender_gap",
      "mcp__frc__list_cohorts",
      "mcp__frc__get_conditional_door_map",
      "mcp__frc__get_rule",
      "mcp__frc__resolve_fha_annual_mip"
    ],
    "ask": [
      "mcp__frc__check_claim_contract",
      "mcp__frc__locate_denominator",
      "mcp__frc__reading_challenge",
      "mcp__frc__verify_reading_proof",
      "mcp__frc__screen_counterparties",
      "mcp__frc__screen_counterparties_quote",
      "mcp__frc__run_cohort"
    ],
    "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 frc. Review it before you use it.

Capability tags behind 1 rule
  • mcp__frc__get_rule 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.financeratecalc/frc-mcp/verify?hash=67746c6d95aa18a089a8887d582f48148b3767ed6b18dd596f11e872eabae4a9" | jq -e '.changed == false'

Pinned hash 67746c6d95aa, from Nov 18, 2026, 13:18 UTC.

Watch

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

https://radar.anlyon.com/servers/io.github.financeratecalc/frc-mcp/changes.xml

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

Badge

Anlyon Radar badge for io.github.financeratecalc/frc-mcp
[![Anlyon Radar](https://radar.anlyon.com/badge/io.github.financeratecalc/frc-mcp.svg)](https://radar.anlyon.com/servers/io.github.financeratecalc/frc-mcp)

Source: MCP Registry · repository. Synced Nov 18, 2026, 13:18 UTC.

frc-mcp server: questions and answers

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

Is the frc-mcp server safe?

No scan can promise that. What its definitions show: 21 tools, 7 of which can change, send or delete something, and 4 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 frc-mcp server delete data?

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

Can the frc-mcp server do things that cannot be undone?

No tool is classed as irreversible: none of them sends, pays, publishes or merges, going by their definitions.

What tools does the frc-mcp server have?

21 tools: 7 write, 14 read. check_claim_contract, locate_denominator, reading_challenge, verify_reading_proof, screen_counterparties, screen_counterparties_quote, run_cohort, get_national_fha_stats and 13 more.

Does the frc-mcp server need sign-in or OAuth?

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

Does the frc-mcp server send data to outside services?

None of its tools sends, posts or publishes, or takes a recipient, channel or URL.

Has the frc-mcp server changed its tools?

Yes, 8 times since Anlyon Radar started tracking it, most recently on Nov 16, 2026. Every change is on this page and in its Atom feed.