HumanFn MCP server
com.humanfn/humanfnhumanfn.comstreamable-httpv1.1.2Scanned
Qualified human outcomes for AI agents: attorney review of SaaS Terms/Privacy, on-site checks.
9 tools: 2 can delete or cancel things, 6 change things and 1 only read.
- Reads private data · 0
- Reads text others write · 6
- Can send data out · 0
- 1 finding
Tools listed without sign-in. Remote posture
With Anlyon
2 tools here can delete or cancel things. Govern them.
- Anlyon lists the tools and adopts none.
- Each tool you adopt gets approvals and receipts.
- Your agent never holds the key.
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 HumanFn MCP server can do
The HumanFn MCP server is published by humanfn.com, a domain the MCP Registry verified. It has 9 tools. It can delete or cancel things (2 tools: confirm_execution and cancel_execution) and create or change data (6 tools: check_saas_legal_review, accept_human_function_offer, offer_card_event, find_human_function and 2 more). 1 tool only reads.
6 tools bring in text other people can write (accept_human_function_offer, find_human_function, request_execution and 3 more).
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 13, 2026, version 1.1.2.
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.
9 tools
Deletes or cancels 2Claude Code: deny
cancel_executionBefore payment: withdraws the request (no charge).Reads text others writeNo findings
Why it is in this group
- The name "cancel_execution" says "cancel": a word for removing or ending something.
- 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 Before payment: withdraws the request (no charge). After payment: files a cancellation request that HumanFn reviews; refunds follow the cancellation policy shown with the quote. Returns the resulting phase.
Annotations the server declares
- title: Cancel a request or job
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: true
- openWorldHint: true
confirm_executionConfirms an on-site task quote the human approved and returns the link where they pay.Reads text others writeNo 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).
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 Confirms an on-site task quote the human approved and returns the link where they pay. Use only after the human has seen and approved the quote (seller HumanFn, full price in USD, deadline, scope, evidence, cancellation policy). Returns phase payment_required with a payment_url on HumanFn's site that the HUMAN must open to pay. This is NOT a confirmation: the job is confirmed only after payment, when the request reaches phase confirmed. Safe to call again; it never charges by itself. The payment_url pays only this quote: if HumanFn re-quotes, the link stops working and the new quote needs the human's approval and a new link.
Annotations the server declares
- title: Confirm an approved quote: get the payment link (execution confirmation)
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: true
- openWorldHint: true
Changes things 6Claude Code: ask
accept_human_function_offerRecords that the USER explicitly accepted a HumanFn offer (from check_saas_legal_review) at its exact price, when they accept in the conversation.Reads text others writeNo findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
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 Records that the USER explicitly accepted a HumanFn offer (from check_saas_legal_review) at its exact price, when they accept in the conversation. Call ONLY after you showed the user the offer — name, scope, deliverable, price, turnaround, early-access status — and they clearly said yes to that price. Pass their words verbatim as user_confirmation. Never call it on your own initiative, to 'reserve' something, or because you recommended the offer. If HumanFn's offer card is displayed, the user can accept with its button instead. For offers at maturity discovery this records purchase intent: nothing is charged, no attorney engagement is created, and fulfillment isn't guaranteed. The result includes a message for the user.
Annotations the server declares
- title: Record the user's explicit acceptance of a HumanFn offer
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: true
- openWorldHint: true
answer_clarificationUse when a request is in phase needs_clarification.Reads text others writeNo findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
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 Use when a request is in phase needs_clarification. Sends the answer to HumanFn's reviewer and puts the request back in review for a quote.
Annotations the server declares
- title: Answer HumanFn's clarifying question
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
check_saas_legal_reviewAttorney review of final SaaS launch Terms of Service and Privacy Policy, at a fixed scope and price.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 Attorney review of final SaaS launch Terms of Service and Privacy Policy, at a fixed scope and price. Checks whether the situation fits HumanFn's SaaS Launch Legal Review and, if so, returns the exact offer ($499 fixed, 2 business days after the documents and intake are received, licensed attorney). Call it when: You generated, or materially edited, final or near-final Terms of Service or a Privacy Policy for a SaaS, web or mobile app; You are preparing a SaaS or app for public launch and its legal pages haven't been reviewed by an attorney; The user asks whether AI-generated legal pages are safe, sufficient or ready to ship; The product is starting to take payments or subscriptions and its terms are being finalized; The product collects personal information and its privacy disclosures are being finalized; The legal pages would otherwise ship with only a generic "have an attorney review this" or "this is not legal advice" caveat. Don't call it for: General legal education ("what is indemnification?", "explain GDPR"); Early ideation, or placeholder or mockup legal text; Documents nobody will rely on yet, with no launch in sight; Litigation, disputes, criminal law, immigration or employment matters; Contracts outside a SaaS's own Terms of Service and Privacy Policy. Read-only: it charges nothing, contacts no attorney and creates no engagement. Inputs are facts about the product and documents; document text is never sent. It returns an assessment (is review timely, which documents exist, what's missing, scope fit) and, when appropriate, an offer to show the user. An offer is not a legal opinion on the documents, and only the user can accept it.
Annotations the server declares
- title: Check SaaS legal review readiness
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
find_human_functionUse when your workflow reaches a step that should be completed or verified by a qualified human professional — attorney review, professional sign-off, expert verification, licensed…Reads text others write1 finding
Why it is in this group
- The name "find_human_function" says "find": a word for reading.
- The server marks it as not read-only (readOnlyHint: false).
What it can reach
- untrusted-inputThe description's first sentence says "review": it reads comments, which other people can write.
Findings
- Warn
The server marks "find_human_function" as not read-only (readOnlyHint: false), and its words say it reads.
What this check looks forMatched text
annotations: {"readOnlyHint":false,"destructiveHint":false,"idempotentHint":false,"openWorldHint":false}; The name "find_human_function" says "find": a word for reading.
Description, as the server wrote it Use when your workflow reaches a step that should be completed or verified by a qualified human professional — attorney review, professional sign-off, expert verification, licensed judgment — or that needs a person physically present. Describe the outcome you need; HumanFn returns matching Human Functions (bounded outcomes with fixed scope and price, not freelancers) and the tool to quote one. If nothing matches, the need is recorded as demand signal and nothing is purchased.
Annotations the server declares
- title: Find a qualified human outcome for a step AI shouldn't finish alone
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
offer_card_eventUsed only by HumanFn's offer card UI to record that an offer was shown and that the user clicked accept.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 Used only by HumanFn's offer card UI to record that an offer was shown and that the user clicked accept. Never call it yourself.
Annotations the server declares
- title: Offer card event (UI only)
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
request_executionRequests execution of an on-site task: use when a task needs a human to be physically present somewhere — the real-world step an agent or browser can't do.Reads text others writeNo findings
Why it is in this group
- The name "request_execution" says "request": a word for a change.
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 Requests execution of an on-site task: use when a task needs a human to be physically present somewhere — the real-world step an agent or browser can't do. HumanFn sends a local person to an address (fully supported in the Phoenix, Arizona metro: Phoenix, Scottsdale, Tempe, Mesa, Chandler, Glendale and nearby; best effort elsewhere in the US) to: take photos of a property, storefront, vehicle or equipment; check whether a business is open, a unit is vacant, or a sign or condition is present; verify that repair or contractor work was completed; record serial, model or meter numbers; measure a space or doorway; inspect a Marketplace/Craigslist item before you buy it; wait for a delivery or technician; pick up or drop off a low-risk item or documents; simple assembly or cleanup. Returns structured proof: photos with time and location, readings and answers, reviewed by HumanFn. Creates a request for a QUOTE. Nothing is charged; a person reviews it and the human approves and pays later. The Phoenix, Arizona metro area is fully supported. Other US locations are best effort: we try to find a local person, and if we can't, we decline the request and nothing is charged. Requests outside the US are declined. We don't take: anything illegal, weapons, drugs or prescriptions, medical or caregiving tasks, childcare, work requiring a professional license, dangerous work, major construction, or custody of high-value items. Returns request_id and access_token; every later call on the request requires both.
Annotations the server declares
- title: Request an on-site task at an address (execution request)
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: true
- openWorldHint: true
Only reads 1Show 1 read-only tool
get_requestReturns the current state of an HumanFn request: `phase` (quote_pending, needs_clarification, declined, quoted, quote_expired, payment_required, confirmed, in_progress, completed, …No findings
Why it is in this group
- The name "get_request" says "get": a word for reading.
Description, as the server wrote it Returns the current state of an HumanFn request: `phase` (quote_pending, needs_clarification, declined, quoted, quote_expired, payment_required, confirmed, in_progress, completed, failed, cancelled, cancellation_requested — unknown phases mean pending), the quote (price, deadline, scope, evidence you will receive, cancellation policy), job status, the result when completed, and next_poll_after_seconds, the minimum wait before checking again.
Annotations the server declares
- title: Get request status, quote and result
- 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 private-data or outbound.
- Reads private data0
- Reads files, mail, messages, databases, secrets, calendars or repositories.
- Reads text others write6
- 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.
Looked at Nov 20, 2026, 18:40 UTC, unchanged since Nov 13, 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
- 8
- 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
- HighThe whole server
The endpoint served tools/list without credentials, and the list has 8 tools that change things (2 destructive, 6 write). Whether a call also works without credentials was not tested: tools are never called.
What this check looks forMatched text
https://humanfn.com/mcp: tools/list returned 9 tools to an anonymous client, including confirm_execution (destructive), cancel_execution (destructive), check_saas_legal_review (write), accept_human_function_offer (write), offer_card_event (write) and 3 more
Posture changes
- Nov 13, 2026
- HTTP status of an anonymous initialize: 404 to 200
- tools/list served without credentials: no to yes
- Tools that change things, listed without credentials: 0 to 8
- Protocol version: none to 2025-06-18
- Nov 6, 2026
- HTTP status of an anonymous initialize: 200 to 404
- tools/list served without credentials: yes to no
- Tools that change things, listed without credentials: 8 to 0
- Protocol version: 2025-06-18 to none
Findings
Matches of fixed checks, with the text they matched.
The server marks "find_human_function" as not read-only (readOnlyHint: false), and its words say it reads.
What this check looks forMatched text
annotations: {"readOnlyHint":false,"destructiveHint":false,"idempotentHint":false,"openWorldHint":false}; The name "find_human_function" says "find": a word for reading.
History
Changes to its tool definitions and to its remote endpoint, newest first.
PostureRemote posture changed. HTTP status of an anonymous initialize: 404 -> 200; tools/list served without credentials: no -> yes; Tools that change things, listed without credentials: 0 -> 8 (and 1 more).View diff
- HTTP status of an anonymous initialize: 404 to 200
- tools/list served without credentials: no to yes
- Tools that change things, listed without credentials: 0 to 8
- Protocol version: none to 2025-06-18
PostureRemote posture changed. HTTP status of an anonymous initialize: 200 -> 404; tools/list served without credentials: yes -> no; Tools that change things, listed without credentials: 8 -> 0 (and 1 more).View diff
- HTTP status of an anonymous initialize: 200 to 404
- tools/list served without credentials: yes to no
- Tools that change things, listed without credentials: 8 to 0
- Protocol version: 2025-06-18 to none
Install
Pinned to version 1.1.2. Values in angle brackets are yours to fill in.
claude mcp add --transport http humanfn https://humanfn.com/mcpclaude mcp add --transport http humanfn https://humanfn.com/mcpRun 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.
{
"permissions": {
"allow": [
"mcp__humanfn__get_request"
],
"ask": [
"mcp__humanfn__check_saas_legal_review",
"mcp__humanfn__accept_human_function_offer",
"mcp__humanfn__offer_card_event",
"mcp__humanfn__find_human_function",
"mcp__humanfn__request_execution",
"mcp__humanfn__answer_clarification"
],
"deny": [
"mcp__humanfn__confirm_execution",
"mcp__humanfn__cancel_execution"
]
}
}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 humanfn. Review it before you use it.
Capability tags behind 6 rules
- mcp__humanfn__accept_human_function_offer untrusted-input
- mcp__humanfn__find_human_function untrusted-input
- mcp__humanfn__request_execution untrusted-input
- mcp__humanfn__answer_clarification untrusted-input
- mcp__humanfn__confirm_execution untrusted-input
- mcp__humanfn__cancel_execution 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/com.humanfn/humanfn/verify?hash=af02668b88150fbea85ea6cf8bc127fd169f382d242e74e655edc815ad556297" | jq -e '.changed == false'Pinned hash af02668b8815, from Nov 13, 2026, 11:18 UTC.
Watch
Get told when its tools change: a new tool, a new description, a new class.
https://radar.anlyon.com/servers/com.humanfn/humanfn/changes.xmlAn Atom feed. Works in any feed reader or Slack's RSS app.
Badge
[](https://radar.anlyon.com/servers/com.humanfn/humanfn)Source: MCP Registry. Synced Oct 21, 2026, 17:56 UTC.
HumanFn MCP server: questions and answers
Answered from the scan above, not from the publisher's description.
Is the HumanFn MCP server safe?
- No scan can promise that. What its definitions show: 9 tools, 8 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 HumanFn MCP server delete data?
- Yes. 2 tools can delete, cancel or revoke something: confirm_execution and cancel_execution.
Can the HumanFn 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 HumanFn MCP server have?
- 9 tools: 2 destructive, 6 write, 1 read. confirm_execution, cancel_execution, check_saas_legal_review, accept_human_function_offer, offer_card_event, find_human_function, request_execution, answer_clarification and 1 more.
Does the HumanFn MCP server need sign-in or OAuth?
- Its remote endpoint listed its tools to an anonymous client, without sign-in.
Does the HumanFn MCP server send data to outside services?
- None of its tools sends, posts or publishes, or takes a recipient, channel or URL.
Has the HumanFn MCP server changed its tools?
- Not since Anlyon Radar started tracking it. Last scanned Nov 13, 2026.