Skip to content

Agent Broker MCP server

dev.hatchloop/agent-brokerhatchloop.devstreamable-httpv0.2.13Scanned

23 MCP tools: compliance, verification, messaging, booking, US contracts. 15 need no key.

23 tools: 5 can delete or cancel things, 1 do something that can't be undone, 10 change things and 7 only read.

Tools: 5 delete, 1 irreversible, 10 write, 7 read.
  • Reads private data · 1
  • Reads text others write · 4
  • Can send data out · 6
  • 1 finding

Tools listed without sign-in. Remote posture

With Anlyon

5 tools here can delete or cancel things and 1 does something that can't be undone. 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 Agent Broker MCP server can do

The Agent Broker MCP server is published by hatchloop.dev, a domain the MCP Registry verified. It has 23 tools. It can delete or cancel things (5 tools: send_message, schedule_appointment, send_transactional_confirmation, escalate_to_human and 1 more), do things that cannot be undone, such as sending, paying or publishing (capture_lead) and create or change data (10 tools: verify_business, handle_inbound, self_test, check_booking_link and 6 more). 7 tools only read.

1 tool reads private data (get_conversation); 4 tools bring in text other people can write (send_message, schedule_appointment, call_business and 1 more); 6 tools can send data out of your system (send_message, capture_lead, send_transactional_confirmation and 3 more). 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 1 finding (annotation-conflict). Each one comes with the text it matched.

Its remote endpoint lists its tools without sign-in.

Last scanned Nov 11, 2026, version 0.2.13.

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.

23 tools

Deletes or cancels 5Claude Code: deny
  • call_business[UNAVAILABLE on this deployment: Voice calling is not enabled on this deployment yet: the outbound phone line has not been verified for placing calls.] Place a conversational voice…Reads text others writeNo findings

    Why it is in this group

    • The name "call_business" says "call": a word for a change.
    • The server marks it destructive (destructiveHint: true).

    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
    [UNAVAILABLE on this deployment: Voice calling is not enabled on this deployment yet: the outbound phone line has not been verified for placing calls.] Place a conversational voice-AI phone call to a business on a consumer's behalf; give a plain-language objective and it navigates the call, extracting the answer. Business-directed (B2B), far less restricted than calling consumers — but the compliance gate still enforces recording consent per jurisdiction. [$0.2/per_call, not charged while no payment rail is on] [async→get_outcome]

    Annotations the server declares

    • title: Call Business
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: true
  • escalate_to_humanHand off an in-flight task to a human operator with a context bundle (transcript, prior actions, identifiers, recommended next step).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 destructive (destructiveHint: true).
    Description, as the server wrote it
    Hand off an in-flight task to a human operator with a context bundle (transcript, prior actions, identifiers, recommended next step). Writes a ticket to the operator queue; sends no notification, promises no response time. [$0.2/per_call, not charged while no payment rail is on] [async→get_outcome] [beta]

    Annotations the server declares

    • title: Escalate To Human
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: false
  • schedule_appointmentCheck availability, book, or cancel via Cal.com.Reads text others writeNo findings

    Why it is in this group

    • The name "schedule_appointment" says "schedule": a word for a change.
    • The server marks it destructive (destructiveHint: true).

    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
    Check availability, book, or cancel via Cal.com. Only completes when the SMB's imported booking link is bound to this deployment's ONE connected Cal.com account; otherwise fails honestly (booking_destination_unmapped) with no charge. No working booking path exists yet for non-Cal.com platforms (voice AI, web form) — those SMBs fail honestly instead of a fabricated confirmation. [from $0.15/call, variable, not charged while no payment rail is on] [async→get_outcome] [limited]

    Annotations the server declares

    • title: Schedule Appointment
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: true
  • send_messageSend a message on behalf of an agent's user or an SMB across WhatsApp (free during launch), SMS, email, or voice, sent immediately, no scheduling.Reads text others writeCan send data outNo findings

    Why it is in this group

    • The name "send_message" says "send": a word for sending, and what is sent cannot be taken back.
    • The input addresses people (input.recipient): what it sends cannot be taken back.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
    • outboundThe name "send_message" says "send": it sends or publishes something to other people.
    Description, as the server wrote it
    Send a message on behalf of an agent's user or an SMB across WhatsApp (free during launch), SMS, email, or voice, sent immediately, no scheduling. Every send passes a non-bypassable gate: TCPA, GDPR, CASL where modeled, else the service's own opt-in default (the Gulf included). Marketing on SMS, voice, WhatsApp and non-US email without recorded opt-in is rejected (compliance_violation receipt). US marketing email follows CAN-SPAM opt-out rules. [from $0.02/call, variable, not charged while no payment rail is on] [async→get_outcome] [Not available on this deployment: SMS, Voice.]

    Annotations the server declares

    • title: Send Message
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: true
  • send_transactional_confirmationIdempotent transactional messages: OTPs, booking confirmations, payment receipts, cancellation notices.Can send data outNo findings

    Why it is in this group

    • The name "send_transactional_confirmation" says "send": a word for sending, and what is sent cannot be taken back.
    • The input addresses people (input.recipient): what it sends cannot be taken back.
    • The server marks it destructive (destructiveHint: true).

    What it can reach

    • outboundThe name "send_transactional_confirmation" says "send": it sends or publishes something to other people.
    Description, as the server wrote it
    Idempotent transactional messages: OTPs, booking confirmations, payment receipts, cancellation notices. Falls back across configured channels; an unconfigured channel fails honestly rather than reporting a delivery that did not happen. [$0.02/per_call, not charged while no payment rail is on] [async→get_outcome] [Not available on this deployment: SMS.]

    Annotations the server declares

    • title: Send Transactional Confirmation
    • readOnlyHint: false
    • destructiveHint: true
    • idempotentHint: false
    • openWorldHint: false
Can't be undone 1Claude Code: ask
  • capture_leadStructured intake of a prospect into an SMB's funnel with validation and deduplication on (smb_id, phone or email).Can send data outNo findings

    Why it is in this group

    • The name "capture_lead" says "capture": a word that usually means money moves, which cannot be taken back.

    What it can reach

    • outboundThe input takes a destination (input.prospect.phone): what the tool writes goes there.
    Description, as the server wrote it
    Structured intake of a prospect into an SMB's funnel with validation and deduplication on (smb_id, phone or email). Stored in AgentBroker's own lead funnel, NOT the business's CRM; the business is not notified. [$0.05/per_call, not charged while no payment rail is on] [async→get_outcome] [beta]

    Annotations the server declares

    • title: Capture Lead
    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: false
    • openWorldHint: false
Changes things 10Claude Code: ask
  • check_complianceFree, instant pre-flight for the compliance gate: runs the SAME TCPA/GDPR/CASL/CAN-SPAM/10DLC checks as send_message and call_business, in preview mode -- nothing is sent, nothing …Can send data outNo 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.

    What it can reach

    • outboundThe input takes a destination (input.channel): what the tool writes goes there.
    Description, as the server wrote it
    Free, instant pre-flight for the compliance gate: runs the SAME TCPA/GDPR/CASL/CAN-SPAM/10DLC checks as send_message and call_business, in preview mode -- nothing is sent, nothing changes. Tells you whether a send would be permitted before you pay, and if not, names the exact rule and how to fix it. De-risks a paid send the way check_booking_link de-risks a paid booking. [free, no key]

    Annotations the server declares

    • title: Check Compliance
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • handle_inboundClassify an inbound message for an SMB and get a suggested_action for YOUR agent to act on.Can send data outNo findings

    Why it is in this group

    • The name "handle_inbound" says "handle": a word for a change.

    What it can reach

    • outboundThe input takes a destination (input.sender.phone): what the tool writes goes there.
    Description, as the server wrote it
    Classify an inbound message for an SMB and get a suggested_action for YOUR agent to act on. English keyword rules (booking request, cancellation, opt-out, inquiry, complaint): treat the intent as a hint. Opt-outs are recorded durably; nothing else is auto-routed or executed. [$0.03/per_call, not charged while no payment rail is on] [async→get_outcome] [beta]

    Annotations the server declares

    • title: Handle Inbound
    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: false
    • openWorldHint: false
  • import_booking_urlTurn a public booking URL into a callable smb_id for send_message and capture_lead.No findings

    Why it is in this group

    • The name "import_booking_url" says "import": a word for a change.
    Description, as the server wrote it
    Turn a public booking URL into a callable smb_id for send_message and capture_lead. But schedule_appointment only completes on Cal.com bound to our one connected account — the other 11 always fail schedule_appointment honestly, uncharged. Detects 12 platforms (Cal.com, Calendly, Doctolib, Booksy, Fresha, OpenTable, Setmore, Square, Acuity, Schedulista, Squarespace, BookMyCity). Idempotent — calling twice returns the same smb_id. [free, requires key] [limited]

    Annotations the server declares

    • title: Import Booking Url
    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • map_trade_restrictionFree cross-border screening of destination and parties: DOES NOT CLASSIFY THE PRODUCT: echoed back, never checked against export-control lists, so a controlled item to an unrestric…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
    Free cross-border screening of destination and parties: DOES NOT CLASSIFY THE PRODUCT: echoed back, never checked against export-control lists, so a controlled item to an unrestricted destination returns 'partial', never 'clear'. Not an export-control clearance -- classify HS/ECCN yourself. Never fabricates a tariff rate, a clear, or a restricted status. [free, no key]

    Annotations the server declares

    • title: Map Trade Restriction
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • mint_keyMint a free-tier API key without email by signing agent_id + timestamp + nonce with the MACHINE_MINT_SECRET (HMAC-SHA256).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.
    Description, as the server wrote it
    Mint a free-tier API key without email by signing agent_id + timestamp + nonce with the MACHINE_MINT_SECRET (HMAC-SHA256). Refused without that secret (invalid_request); not_configured if the operator has none set. No secret: use POST /keys/request. [free, no key] [limited]

    Annotations the server declares

    • title: Mint Key
    • readOnlyHint: false
    • destructiveHint: false
    • idempotentHint: false
    • openWorldHint: false
  • screen_sanctionsFree screening of a name/entity (Latin or Arabic script) against sanctions lists: OFAC SDN (US Treasury), EU consolidated list (European Commission), UK Sanctions List (FCDO).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
    Free screening of a name/entity (Latin or Arabic script) against sanctions lists: OFAC SDN (US Treasury), EU consolidated list (European Commission), UK Sanctions List (FCDO). THE UN CONSOLIDATED LIST IS NOT SCREENED -- no licence permits redistribution. Never fabricates a match or clear: check lists_screened; reason_code 'partial_screening' (never 'clear') if any source failed. [free, no key]

    Annotations the server declares

    • title: Screen Sanctions
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • self_testService health probe: runs 6 internal checks and reports how many passed.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
    Service health probe: runs 6 internal checks and reports how many passed. Confirms the server is up and responding - it does NOT probe each tool individually. Use to verify connectivity before production use. [free, no key]

    Annotations the server declares

    • title: Self Test
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • verify_businessLook up what we know about a business in our supply network: contact channels, capabilities, last verification.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 what we know about a business in our supply network: contact channels, capabilities, last verification. A DIRECTORY LOOKUP - it does not contact the business. Supply-network ids only, not osm:... ids. [free, no key] [limited]

    Annotations the server declares

    • title: Verify Business
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • verify_company_recordFree, live lookup of a company official registry record.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
    Free, live lookup of a company official registry record. Queries the GLEIF global LEI registry (primary, 2.6 million legal entities worldwide) and SEC EDGAR (US public companies) to return the official legal name, LEI, entity status, jurisdiction, registered address, and registry authority. Never fabricates: if the company is not found in these free registries, returns an honest not_found with the sources that were queried. [free, no key]

    Annotations the server declares

    • title: Verify Company Record
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
Only reads 7Show 7 read-only tools
  • check_quotaReturns the caller's current quota state: how many operations have been used today, how many remain, the daily limit, the tier (free / credits / unlimited), and when the quota rese…No findings

    Why it is in this group

    • The description's first verb is "return": a word for reading.
    Description, as the server wrote it
    Returns the caller's current quota state: how many operations have been used today, how many remain, the daily limit, the tier (free / credits / unlimited), and when the quota resets. Read-only and free. Works with any valid X-Agent-Identity token; returns tier='anonymous' for keyless callers with the anonymous daily limits. [free, no key]

    Annotations the server declares

    • title: Check Quota
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • find_businessFind real businesses near a place, sorted by distance (nearest of those examined; see result.search).No findings

    Why it is in this group

    • The name "find_business" says "find": a word for reading.
    Description, as the server wrote it
    Find real businesses near a place, sorted by distance (nearest of those examined; see result.search). Data: OpenStreetMap (ODbL, (c) OSM contributors); phone, website, hours only when mapped. Coverage varies; community-mapped, NOT verified by us; source=openstreetmap rows cannot be booked. Answers within about 5 s: status partial (search_in_progress) means repeat the call. If OSM is down: retriable 'temporarily unavailable', never invented rows. [free, no key] [beta]

    Annotations the server declares

    • title: Find Business
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_conversationRead a two-way conversation thread you started: its state, every message exchanged, and how many replies the business has sent.Reads private dataReads text others writeNo findings

    Why it is in this group

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

    What it can reach

    • private-dataThe name "get_conversation" says "get" and "conversation": it reads messages.
    • private-dataThe description's first sentence says "read" and "conversation": it reads messages.
    • untrusted-inputThe name "get_conversation" says "get" and "conversation": it reads messages, which other people can write.
    • untrusted-inputThe description's first sentence says "read" and "conversation": it reads messages, which other people can write.
    Description, as the server wrote it
    Read a two-way conversation thread you started: its state, every message exchanged, and how many replies the business has sent. [free, requires key]

    Annotations the server declares

    • title: Get Conversation
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: false
    • openWorldHint: false
  • get_outcomeRetrieve the final OutcomeReceipt for a completed operation.No findings

    Why it is in this group

    • The name "get_outcome" says "get": a word for reading.
    Description, as the server wrote it
    Retrieve the final OutcomeReceipt for a completed operation. [free, no key]

    Annotations the server declares

    • title: Get Outcome
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • get_statusQuery the current state of any in-flight async operation by operation_id.No findings

    Why it is in this group

    • The name "get_status" says "get": a word for reading.
    Description, as the server wrote it
    Query the current state of any in-flight async operation by operation_id. [free, no key]

    Annotations the server declares

    • title: Get Status
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • lookup_us_contractsSearch US federal contract awards by company (recipient) name using the free USASpending.gov public API.No findings

    Why it is in this group

    • The name "lookup_us_contracts" says "lookup": a word for reading.
    Description, as the server wrote it
    Search US federal contract awards by company (recipient) name using the free USASpending.gov public API. Returns the top awards sorted by dollar amount: award ID, recipient name, award amount USD, awarding agency, NAICS code/description, and period of performance. Zero upstream cost, no key required. Data covers 2020-2026 procurement contracts. [free, no key]

    Annotations the server declares

    • title: Lookup Us Contracts
    • readOnlyHint: true
    • destructiveHint: false
    • idempotentHint: true
    • openWorldHint: false
  • preview_costReturn an expected cost estimate, latency estimate, and success-probability estimate for a proposed call before execution.No findings

    Why it is in this group

    • The name "preview_cost" says "preview": a word for reading.
    Description, as the server wrote it
    Return an expected cost estimate, latency estimate, and success-probability estimate for a proposed call before execution. Returns the exact price when it is fixed, and a min/max range when the cost depends on channel or outcome. It does not promise an accuracy percentage - check cost_range. [free, no key]

    Annotations the server declares

    • title: Preview Cost
    • 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.

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

Reads private data1
Reads files, mail, messages, databases, secrets, calendars or repositories.
Reads text others write4
Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
Can send data out6
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__agent-broker__get_conversation"
    ]
  }
}

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, 18:52 UTC, unchanged since Oct 22, 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
16
Protocol version
2025-06-18
WWW-Authenticate on 401
not observed
resource_metadata in WWW-Authenticate
not observed
RFC 9728 protected resource metadata
https://hatchloop.dev/.well-known/oauth-protected-resource/mcp/agent-broker
Authorization servers
https://api.hatchloop.dev
Legacy HTTP+SSE transport
no
Dynamic client registration
yes
Client ID metadata documents
yes
  • HighThe whole server

    The endpoint served tools/list without credentials, and the list has 16 tools that change things (5 destructive, 1 irreversible, 10 write). Whether a call also works without credentials was not tested: tools are never called.

    Matched text

    https://hatchloop.dev/mcp/agent-broker: tools/list returned 23 tools to an anonymous client, including send_message (destructive), schedule_appointment (destructive), send_transactional_confirmation (destructive), escalate_to_human (destructive), call_business (destructive) and 11 more

    What this check looks for

Findings

Matches of fixed checks, with the text they matched.

1 warn
  • The server marks "check_booking_link" 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 "check_booking_link" says "link": a word for a change.

    What this check looks for

History

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

No change seen yet. Its tools are rescanned every day, and a remote endpoint is looked at once a week.

Install

Pinned to version 0.2.13. Values in angle brackets are yours to fill in.

claude mcp add --transport http agent-broker https://hatchloop.dev/mcp/agent-broker
claude mcp add --transport http agent-broker https://hatchloop.dev/mcp/agent-broker

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__agent-broker__find_business",
      "mcp__agent-broker__get_status",
      "mcp__agent-broker__get_outcome",
      "mcp__agent-broker__preview_cost",
      "mcp__agent-broker__check_quota",
      "mcp__agent-broker__get_conversation",
      "mcp__agent-broker__lookup_us_contracts"
    ],
    "ask": [
      "mcp__agent-broker__verify_business",
      "mcp__agent-broker__capture_lead",
      "mcp__agent-broker__handle_inbound",
      "mcp__agent-broker__self_test",
      "mcp__agent-broker__check_booking_link",
      "mcp__agent-broker__import_booking_url",
      "mcp__agent-broker__check_compliance",
      "mcp__agent-broker__verify_company_record",
      "mcp__agent-broker__screen_sanctions",
      "mcp__agent-broker__map_trade_restriction",
      "mcp__agent-broker__mint_key"
    ],
    "deny": [
      "mcp__agent-broker__send_message",
      "mcp__agent-broker__schedule_appointment",
      "mcp__agent-broker__send_transactional_confirmation",
      "mcp__agent-broker__escalate_to_human",
      "mcp__agent-broker__call_business"
    ]
  }
}

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 agent-broker. Review it before you use it.

Capability tags behind 9 rules
  • mcp__agent-broker__send_message untrusted-input, outbound
  • mcp__agent-broker__capture_lead outbound
  • mcp__agent-broker__schedule_appointment untrusted-input
  • mcp__agent-broker__send_transactional_confirmation outbound
  • mcp__agent-broker__handle_inbound outbound
  • mcp__agent-broker__check_booking_link outbound
  • mcp__agent-broker__call_business untrusted-input
  • mcp__agent-broker__check_compliance outbound
  • mcp__agent-broker__get_conversation private-data, untrusted-input

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/dev.hatchloop/agent-broker/verify?hash=a65cea75db0ff712124c8a11281c74cdbe173be392123aa82c2a2b83982fcd34" | jq -e '.changed == false'

Pinned hash a65cea75db0f, from Nov 11, 2026, 08:17 UTC.

Watch

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

https://radar.anlyon.com/servers/dev.hatchloop/agent-broker/changes.xml

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

Badge

Anlyon Radar badge for dev.hatchloop/agent-broker
[![Anlyon Radar](https://radar.anlyon.com/badge/dev.hatchloop/agent-broker.svg)](https://radar.anlyon.com/servers/dev.hatchloop/agent-broker)

Source: MCP Registry · repository. Synced Oct 21, 2026, 18:11 UTC.

Agent Broker MCP server: questions and answers

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

Is the Agent Broker MCP server safe?

No scan can promise that. What its definitions show: 23 tools, 16 of which can change, send or delete something, and 1 finding from fixed checks. Allow the reads, ask before each change, and deny the deletes; the permissions block on this page does that.

Can the Agent Broker MCP server delete data?

Yes. 5 tools can delete, cancel or revoke something: send_message, schedule_appointment, send_transactional_confirmation, escalate_to_human and call_business.

Can the Agent Broker MCP server do things that cannot be undone?

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

What tools does the Agent Broker MCP server have?

23 tools: 5 destructive, 1 irreversible, 10 write, 7 read. send_message, schedule_appointment, send_transactional_confirmation, escalate_to_human, call_business, capture_lead, verify_business, handle_inbound and 15 more.

Does the Agent Broker MCP server need sign-in or OAuth?

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

Does the Agent Broker MCP server send data to outside services?

6 tools can carry data out (it sends, posts or publishes, or takes a recipient, channel or URL): send_message, capture_lead, send_transactional_confirmation, handle_inbound, check_booking_link and check_compliance.

Has the Agent Broker MCP server changed its tools?

Not since Anlyon Radar started tracking it. Last scanned Nov 11, 2026.