person.tools MCP server
tools.person/ask-peopleperson.toolsstreamable-httpv0.1.0Scanned
Ask people for a reaction a model can't give; get each answer and what their page observed.
6 tools: 3 can delete or cancel things, 2 change things and 1 only read.
- Reads private data · 0
- Reads text others write · 3
- Can send data out · 1
- 1 finding
Tools listed without sign-in. Remote posture
Anlyon alternative
5 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 person.tools MCP server can do
The person.tools MCP server is published by person.tools, a domain the MCP Registry verified. It has 6 tools. It can delete or cancel things (3 tools: ask_people, ask_people_cancel and ask_people_send_outcome) and create or change data (2 tools: ask_people_quote and ask_people_panel). 1 tool only reads.
3 tools bring in text other people can write (ask_people, ask_people_send_outcome and ask_people_panel); 1 tool can send data out of your system (ask_people_send_outcome).
The fixed checks matched 1 finding (long-description). Each one comes with the text it matched.
Its remote endpoint lists its tools without sign-in.
Last scanned Nov 15, 2026, version 0.1.0.
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.
6 tools
Deletes or cancels 3Claude Code: deny
ask_peopleOpens a call: one short task to several people.Reads text others write1 finding
Why it is in this group
- The description's first verb is "open": 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.
Findings
- Info
The description is 3,386 characters long, over the 2,000 this check looks for.
What this check looks forMatched text
description ends: "...ld_play_again":"boolean"},"max_total_usd":"0.00","idempotency_key":"first-try-1"}"
Description, as the server wrote it Opens a call: one short task to several people. During the preview a test key or no key reaches simulated people: their answers are made up and marked simulated. Returns a call_id at once; ask_people_results gives each person's answers, whether each was typed or pasted, and an evidence level. Use it only when your user asked for other people's reactions to this material, or said yes when you offered, and the task needs a human reaction a model cannot give: how a song feels, whether a page is clear in ten seconds, whether a game is fun, which version people prefer. Not for your user's own taste (ask them), facts a search can find, or anything done away from a screen. Who answers: pool defaults to "simulated". Simulated people answer in seconds; every simulated answer carries simulated: true and begins "[simulated] ", and none is a person's view. Real people are rung only on pool "panel:<panel_id>", a panel the account owner invited, in minutes when they are available and never in their quiet hours (9 PM to 8 AM their time by default). Each person is told what is recorded before they take the call. Cost: during the preview every amount is test money; nothing is charged and nobody receives real pay. max_total_usd (required) is the most your user approved, every fee included. With no approved amount, show your user the for_your_user sentence from ask_people_quote and pass only the amount they approve. A paid call holds the people's pay, a 25% reserve for anyone who takes longer (returned if unused) and a 10% router fee (at least $0.25), each on its own line. A simulated call or a panel at zero pay costs nothing. Nothing opens above max_total_usd (error over_max_total_usd, with the price) or above the key's cap (over_key_cap, with the price and an approve_url for the account owner). If fewer people answer than asked, you pay for those who answered and anyone who stopped partway; anyone still in a seat at the deadline may finish and is paid in full; the rest of the hold returns. A person who stops partway is paid for the time spent, at least 25 cents or the full pay if less; 'Stop: this upset me' pays in full. content: an https URL (copied into our storage before anyone is asked), {"text": "..."}, or for compare a list of 2 to 6 of these; not used for kind answer. Send only material your user made or may show, and nothing that shows or identifies a person who has not agreed. Ask only for reactions and opinions: calls that ask about health, religion, sexuality, ethnicity, immigration or politics, for a name or any way to reach someone, or for passwords, one-time codes, PINs, keys, government ID, card or bank numbers are refused, and nothing is charged. Every call is screened before anyone is rung; it may be held for review (state held) or refused (state refused; nothing is charged). Each default used comes back in defaults_used. Look again after next_check_after_seconds. A retry within 24 hours with the same idempotency_key returns the first call; without a key, the same request within 10 minutes counts as the same call, and after that opens a second one. Example: {"ask":"Listen once to the end. Then say what you felt, in your own words.","kind":"listen","content":"https://example.com/demo-track.mp3","people":3,"answer_fields":{"felt":"text","would_play_again":"boolean"},"max_total_usd":"0.00","idempotency_key":"first-try-1"}
Annotations the server declares
- title: Ask people
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: false
- openWorldHint: true
ask_people_cancelStops a call.No findings
Why it is in this group
- The name "ask_people_cancel" says "cancel": a word for removing or ending something.
- The description says "cannot be undone".
- The server marks it destructive (destructiveHint: true).
Description, as the server wrote it Stops a call. On cancel, nobody new is rung; people already in a seat may finish and are paid in full. Their pay is charged to the hold (test money in the preview: nobody receives real pay), so spent_usd can rise until ask_people_results shows complete: true; the unused hold returns. Returns the call's state and the money so far. Repeating it changes nothing; a cancel cannot be undone. Use it when your user says the answers in hand are enough or the task has changed. A call closes by itself at its deadline. Example: {"call_id": "call_..."}
Annotations the server declares
- title: Stop a call to people
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: true
- openWorldHint: false
ask_people_send_outcomeSends the people who answered a call one line (at most 280 characters, and up to three lines a call) saying what you did with their answers, for example "Cut the intro from 40 to 1…Reads text others writeCan send data outNo findings
Why it is in this group
- The name "ask_people_send_outcome" says "send": a word for sending, and what is sent 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 "ask_people_send_outcome" says "send": it sends or publishes something to other people.
Description, as the server wrote it Sends the people who answered a call one line (at most 280 characters, and up to three lines a call) saying what you did with their answers, for example "Cut the intro from 40 to 12 seconds; most of you drifted there." The line is reviewed the way a call is and may be refused; if allowed, it appears on each answering person's receipt and they are notified. Costs nothing. Returns whether the line was allowed and how many people it reached. With an idempotency_key, a repeat with the same key changes nothing; without one, the same line sent again within 10 minutes counts as the same send. Use it, with your user's agreement, once the answers have changed the work. Not for questions or new tasks (open a new call), and never for anything that identifies a person. Example: {"call_id": "call_...", "text": "Cut the intro from 40 to 12 seconds; most of you drifted there.", "idempotency_key": "outcome-1"}
Annotations the server declares
- title: Tell people what their answers changed
- readOnlyHint: false
- destructiveHint: true
- idempotentHint: false
- openWorldHint: true
Changes things 2Claude Code: ask
ask_people_panelMakes a panel, a named group of people the account owner invites, and returns its panel_id and a join link for your user to share.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 Makes a panel, a named group of people the account owner invites, and returns its panel_id and a join link for your user to share. Each person joins on their own phone, states that they are an adult, is told what is recorded while they answer (how much played, whether the page stayed in front, pauses and replays, typed or pasted), chooses when they can be rung, and can leave at any time. Calls with pool "panel:<panel_id>" then reach only them, and cost nothing at zero pay. On a panel of fewer than five, both sides are told that answers cannot be anonymous. Use it only when your user asks to set up a group of people they will invite themselves; nobody is added without joining. Needs a key issued for real calls; a test key gets an error saying so. A retry with the same idempotency_key returns the same panel; without one, a retry makes a second panel. Example: {"name": "album-listeners", "idempotency_key": "panel-1"}
Annotations the server declares
- title: Make a panel of invited people
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
ask_people_quotePrices a call without opening it: nobody is rung and nothing is spent.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 Prices a call without opening it: nobody is rung and nothing is spent. Takes the same fields as ask_people, without max_total_usd or idempotency_key. Nothing else is needed first: leave content out until your user gives you the material (the quote is priced from minutes), and leave pay out unless your user named one (left out, it is 0.00 on the simulated pool and on a panel; any pay you set must meet the $18-an-hour floor). Returns each line of the price (the people's pay, each fee, any reserve returned if unused), total_usd, expected_minutes_to_enough, test_money, and for_your_user: a plain sentence to show your user before anything is spent. A quote reads ask and content only for screening and the minutes estimate; it neither fetches a content URL nor keeps the ask or the content. Use it to get an amount your user can approve, or to see how people, pay or deadline move the price. Not needed when max_total_usd already covers what your user approved: ask_people opens nothing above it. Example: {"ask":"Which headline makes you want to read on, and why?","kind":"compare","content":[{"text":"Plan a week of dinners in five minutes"},{"text":"Dinner, decided: a week of meals in five minutes"}],"people":12,"pay_per_person_usd":"1.00","minutes":2,"pool":"simulated"}
Annotations the server declares
- title: Price a call to people
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
Only reads 1Show 1 read-only tool
ask_people_resultsReturns a call's answers so far, given the call_id from ask_people: first a summary (the state, a tally of each answer field, how many responses reached each evidence level), then …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 a call's answers so far, given the call_id from ask_people: first a summary (the state, a tally of each answer field, how many responses reached each evidence level), then one entry per person with their answers and whether each was typed or pasted, then the money (held_usd, spent_usd, returned_usd, and test_money, true throughout the preview: nobody receives real pay). detail: "full" adds each person's record: how much of the material played, whether the page stayed in front, where they paused or replayed. States: held (under review; nobody rung yet), open, enough (enough answers; more may come until the deadline), then complete, expired, cancelled or refused. Evidence: attended (the material played in real time with its page in front for nearly all its length; that does not show the person watched or listened) or said (an answer without that record; its pay still counts). No level shows that an answer was formed without help. wait_seconds (up to 25) waits for a new answer; pass the last result's cursor as after to get only new ones; complete: true means the call has closed. Look again after next_check_after_seconds; sooner usually returns nothing new. Simulated answers carry simulated: true and begin "[simulated] ". Example: {"call_id": "call_...", "after": "<cursor>", "wait_seconds": 25}
Annotations the server declares
- title: Get answers from a call to people
- 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.
- Reads private data0
- Reads files, mail, messages, databases, secrets, calendars or repositories.
- Reads text others write3
- Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
- Can send data out1
- Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.
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 15, 2026, 13:20 UTC, unchanged since Nov 15, 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
- 5
- 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 5 tools that change things (3 destructive, 2 write). Whether a call also works without credentials was not tested: tools are never called.
What this check looks forMatched text
https://api.person.tools/mcp: tools/list returned 6 tools to an anonymous client, including ask_people (destructive), ask_people_cancel (destructive), ask_people_send_outcome (destructive), ask_people_quote (write), ask_people_panel (write)
Findings
Matches of fixed checks, with the text they matched.
- Infoask_people
The description is 3,386 characters long, over the 2,000 this check looks for.
What this check looks forMatched text
description ends: "...ld_play_again":"boolean"},"max_total_usd":"0.00","idempotency_key":"first-try-1"}"
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 0.1.0. Values in angle brackets are yours to fill in.
claude mcp add --transport http ask-people https://api.person.tools/mcpclaude mcp add --transport http ask-people https://api.person.tools/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__ask-people__ask_people_results"
],
"ask": [
"mcp__ask-people__ask_people_quote",
"mcp__ask-people__ask_people_panel"
],
"deny": [
"mcp__ask-people__ask_people",
"mcp__ask-people__ask_people_cancel",
"mcp__ask-people__ask_people_send_outcome"
]
}
}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 ask-people. Review it before you use it.
Capability tags behind 3 rules
- mcp__ask-people__ask_people untrusted-input
- mcp__ask-people__ask_people_send_outcome untrusted-input, outbound
- mcp__ask-people__ask_people_panel 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/tools.person/ask-people/verify?hash=461422e221767f997ec36e0f9e3164b4e91f09558fb9bab3126613918af0f3c4" | jq -e '.changed == false'Pinned hash 461422e22176, from Nov 15, 2026, 05:18 UTC.
Watch
Get told when its tools change: a new tool, a new description, a new class.
https://radar.anlyon.com/servers/tools.person/ask-people/changes.xmlAn Atom feed. Works in any feed reader or Slack's RSS app.
Badge
[](https://radar.anlyon.com/servers/tools.person/ask-people)Source: MCP Registry · repository. Synced Nov 15, 2026, 10:18 UTC.
person.tools MCP server: questions and answers
Answered from the scan above, not from the publisher's description.
Is the person.tools MCP server safe?
- No scan can promise that. What its definitions show: 6 tools, 5 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 person.tools MCP server delete data?
- Yes. 3 tools can delete, cancel or revoke something: ask_people, ask_people_cancel and ask_people_send_outcome.
Can the person.tools 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 person.tools MCP server have?
- 6 tools: 3 destructive, 2 write, 1 read. ask_people, ask_people_cancel, ask_people_send_outcome, ask_people_quote, ask_people_panel and ask_people_results.
Does the person.tools MCP server need sign-in or OAuth?
- Its remote endpoint listed its tools to an anonymous client, without sign-in.
Does the person.tools MCP server send data to outside services?
- 1 tool can carry data out (it sends, posts or publishes, or takes a recipient, channel or URL): ask_people_send_outcome.
Has the person.tools MCP server changed its tools?
- Not since Anlyon Radar started tracking it. Last scanned Nov 15, 2026.