Human For AI MCP server
dev.humanforai/humanforaihumanforai.devstreamable-httpv0.1.4Scanned
Hire a real human for real-world verification, product testing, AI output review, and errands.
6 tools: 2 do something that can't be undone, 1 change things and 3 only read.
- Reads private data · 1
- Reads text others write · 6
- Can send data out · 2
- No findings
Tools listed without sign-in. Remote posture
Anlyon alternative
3 of these tools change something. Run those as governed actions.
Declare what an action changes, and Anlyon puts it under a shared limit, binds every approval to the exact request, and grades the outcome. It works on any HTTPS API you can describe as an action.
Anlyon Radar does not run, proxy or govern this MCP server. A tool your own code calls directly is not routed through Anlyon and is not governed by it.
Import an OpenAPI spec →What the Human For AI MCP server can do
The Human For AI MCP server is published by humanforai.dev, a domain the MCP Registry verified. It has 6 tools. It can do things that cannot be undone, such as sending, paying or publishing (message_human_operator and reply_in_message_thread) and create or change data (1 tool: submit_human_task), but cannot delete data. 3 tools only read.
1 tool reads private data (check_message_thread); 6 tools bring in text other people can write (get_human_services, submit_human_task, check_task_status and 3 more); 2 tools can send data out of your system (message_human_operator and reply_in_message_thread). On its own it has all three parts of a prompt-injection exfiltration path: private data, text others write, and a way out.
None of the fixed checks matched anything in its tool definitions. That is not a safety guarantee.
Its remote endpoint lists its tools without sign-in.
Last scanned Nov 11, 2026, version 0.1.4.
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.
6 tools
Can't be undone 2Claude Code: ask
message_human_operatorSend a free-form message to the human operator — questions, scoping, custom or recurring projects, anything that is not yet a ready-made task.Reads text others writeCan send data outNo findings
Why it is in this group
- The description's first verb is "send": a word for sending, and what is sent 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 description's first verb is "send": it sends or publishes something to other people.
Description, as the server wrote it Send a free-form message to the human operator — questions, scoping, custom or recurring projects, anything that is not yet a ready-made task. reply_to is REQUIRED: an email a human can read (MX-checked), or an https URL to receive the reply as a signed webhook push. The response also carries thread_url + access_token — every message is a pollable thread, so you can read the reply with check_message_thread even without a mailbox. Keep the token: it is shown only once.
Annotations the server declares
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
reply_in_message_threadAdd a follow-up to a thread you opened — answer a question the operator asked, add detail, correct yourself, or withdraw the request.Reads text others writeCan send data outNo findings
Why it is in this group
- The name "reply_in_message_thread" says "reply": a word for sending, and what is sent 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 "reply_in_message_thread" says "reply": it sends or publishes something to other people.
Description, as the server wrote it Add a follow-up to a thread you opened — answer a question the operator asked, add detail, correct yourself, or withdraw the request. Prefer this over sending a brand-new message about the same subject. Needs the message_id and access_token from the submission response.
Annotations the server declares
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
Changes things 1Claude Code: ask
submit_human_taskSubmit a task for the human operator to perform in the real world.Reads text others writeNo findings
Why it is in this group
- The name "submit_human_task" says "submit": 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 Submit a task for the human operator to perform in the real world. Returns a task_id immediately; the human reviews every task before accepting it (this is not instant execution). The operator is push-notified on submission; check_task_status shows seen_by_operator_at once a human has seen the task. Free during the pilot. contact_email must be a real mailbox (MX-checked) — it is how the deliverable reaches you. No mailbox? Set delivery to 'status_poll' instead: the deliverable arrives as text in operator_notes via check_task_status (limited to 1 such task per client per day). In hosts that support MCP Apps the result also renders as a task status card with a Refresh button; the JSON result carries the same data.
Annotations the server declares
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
Only reads 3Show 3 read-only tools
check_message_threadRead the thread for a message you sent: the original text, every reply oldest-first, and whether the operator has answered.Reads private dataReads text others writeNo findings
Why it is in this group
- The description's first verb is "read": a word for reading.
What it can reach
- private-dataThe name "check_message_thread" says "message": it reads messages.
- private-dataThe description's first sentence says "read" and "thread": it reads messages.
- untrusted-inputThe name "check_message_thread" says "message": it reads messages, which other people can write.
- untrusted-inputThe description's first sentence says "read" and "thread": it reads messages, which other people can write.
Description, as the server wrote it Read the thread for a message you sent: the original text, every reply oldest-first, and whether the operator has answered. Needs the message_id and the access_token from the submission response. An empty replies list means no answer yet — the operator works at human speed, so poll occasionally rather than in a loop.
Annotations the server declares
- readOnlyHint: true
check_task_statusLook up a submitted task by its task_id.Reads text others writeNo findings
Why it is in this group
- The name "check_task_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 Look up a submitted task by its task_id. Returns current status (submitted → accepted → delivered, or rejected), status history with timestamps, seen_by_operator_at (the moment a human actually saw the task — usually well before the first status change), eta (operator-set delivery estimate, once accepted), and any operator notes. Once delivered, the response also carries receipt (a signed JWS binding the deliverable's sha256 to the lifecycle timestamps) and deliverable_sha256 — verify offline against https://humanforai.dev/.well-known/jwks.json. In hosts that support MCP Apps the result also renders as a task status card (timeline, seen-by-human moment, ETA, operator notes, receipt) with a Refresh button; the JSON result carries the same data.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: true
get_human_servicesFetch the Human For AI manifest: available services, operator profile (location, languages, working hours), response times, accepted and rejected task types, and trust & safety pol…Reads text others writeNo findings
Why it is in this group
- The name "get_human_services" says "get": 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 Fetch the Human For AI manifest: available services, operator profile (location, languages, working hours), response times, accepted and rejected task types, and trust & safety policy. Call this first to decide whether and how to hire the human. The catalog is examples, not limits — unlisted needs are welcome as custom_human_in_the_loop.
Annotations the server declares
- readOnlyHint: true
- openWorldHint: true
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.
12 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 write6
- Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
- Can send data out2
- Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.
{
"permissions": {
"deny": [
"mcp__humanforai__check_message_thread"
]
}
}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, 18:52 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
- 3
- Protocol version
- 2025-06-18
- WWW-Authenticate on 401
- not observed
- resource_metadata in WWW-Authenticate
- not observed
- RFC 9728 protected resource metadata
- https://humanforai.dev/.well-known/oauth-protected-resource
- 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 3 tools that change things (2 irreversible, 1 write). Whether a call also works without credentials was not tested: tools are never called.
What this check looks forMatched text
https://humanforai.dev/mcp: tools/list returned 6 tools to an anonymous client, including message_human_operator (irreversible), reply_in_message_thread (irreversible), submit_human_task (write)
Findings
Matches of fixed checks, with the text they matched.
No check matched anything in these tool definitions. That is not a safety guarantee.
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 0.1.4. Values in angle brackets are yours to fill in.
claude mcp add --transport http humanforai https://humanforai.dev/mcpclaude mcp add --transport http humanforai https://humanforai.dev/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__humanforai__get_human_services",
"mcp__humanforai__check_task_status",
"mcp__humanforai__check_message_thread"
],
"ask": [
"mcp__humanforai__submit_human_task",
"mcp__humanforai__message_human_operator",
"mcp__humanforai__reply_in_message_thread"
],
"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 humanforai. Review it before you use it.
Capability tags behind 6 rules
- mcp__humanforai__get_human_services untrusted-input
- mcp__humanforai__submit_human_task untrusted-input
- mcp__humanforai__check_task_status untrusted-input
- mcp__humanforai__message_human_operator untrusted-input, outbound
- mcp__humanforai__check_message_thread private-data, untrusted-input
- mcp__humanforai__reply_in_message_thread untrusted-input, outbound
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.humanforai/humanforai/verify?hash=7fa0747d6c945cb934635782531e188b85101436cc0c4956d9a04b37460dd6bd" | jq -e '.changed == false'Pinned hash 7fa0747d6c94, from Nov 11, 2026, 07:18 UTC.
Watch
Get told when its tools change: a new tool, a new description, a new class.
https://radar.anlyon.com/servers/dev.humanforai/humanforai/changes.xmlAn Atom feed. Works in any feed reader or Slack's RSS app.
Badge
[](https://radar.anlyon.com/servers/dev.humanforai/humanforai)Source: MCP Registry · repository. Synced Oct 21, 2026, 18:11 UTC.
Human For AI MCP server: questions and answers
Answered from the scan above, not from the publisher's description.
Is the Human For AI MCP server safe?
- No scan can promise that. What its definitions show: 6 tools, 3 of which can change, send or delete something, and 0 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 Human For AI MCP server delete data?
- No tool is classed as destructive: going by their definitions, none of them deletes, cancels or revokes anything.
Can the Human For AI MCP server do things that cannot be undone?
- Yes. 2 tools can send, pay, publish or merge: message_human_operator and reply_in_message_thread.
What tools does the Human For AI MCP server have?
- 6 tools: 2 irreversible, 1 write, 3 read. message_human_operator, reply_in_message_thread, submit_human_task, get_human_services, check_task_status and check_message_thread.
Does the Human For AI MCP server need sign-in or OAuth?
- Its remote endpoint listed its tools to an anonymous client, without sign-in.
Does the Human For AI MCP server send data to outside services?
- 2 tools can carry data out (it sends, posts or publishes, or takes a recipient, channel or URL): message_human_operator and reply_in_message_thread.
Has the Human For AI MCP server changed its tools?
- Not since Anlyon Radar started tracking it. Last scanned Nov 11, 2026.