Screwtape.ai MCP server
ai.screwtape/screwtapescrewtape.aistreamable-httpv1.4.0Scanned
Spend guardrails for AI agents: check payments, track budgets, catch duplicates, sign receipts.
9 tools: 3 do something that can't be undone and 6 change things.
- Reads private data · 0
- Reads text others write · 5
- Can send data out · 1
- 1 finding
Tools listed without sign-in. Remote posture
With Anlyon
3 tools here do 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.
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 Screwtape.ai MCP server can do
The Screwtape.ai MCP server is published by screwtape.ai, a domain the MCP Registry verified. It has 9 tools. It can do things that cannot be undone, such as sending, paying or publishing (budget_ledger, payment_check and publish_receipt) and create or change data (6 tools: action_receipt, judge, check_duplicate_invoice, consensus and 2 more), but cannot delete data. None of its tools only reads.
5 tools bring in text other people can write (payment_check, judge, check_duplicate_invoice and 2 more); 1 tool can send data out of your system (publish_receipt).
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, and its tools changed 6 times since Anlyon Radar first saw them.
Last scanned Nov 10, 2026, version 1.4.0.
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.
9 tools
Can't be undone 3Claude Code: ask
budget_ledgerBudget Ledger: atomically reserves reported USD spend across a run and its sub-agents.No findings
Why it is in this group
- The input carries an amount (input.amount) and a currency, so it moves money.
Description, as the server wrote it Budget Ledger: atomically reserves reported USD spend across a run and its sub-agents. Returns allow/warn/block, remaining budget, retry storms and spend velocity. Call judge first, then reserve immediately before spending; honor block. This is cooperative accounting, not payment enforcement. Root cap and parent are immutable. Children inherit the root cap (send the same cap). event_id is the authoritative idempotency key; a custom HTTP key cannot override it. New event_id per attempted action; same event_id and body replays without reserving twice. No refunds or cap resets. Warn reservations count as spend. Limits: USD, six decimals, cap at most 1000000. Three matching actions in 60 seconds allowed; fourth blocked. Velocity warning at ten attempts/minute or more than 25% of cap/minute; 80% usage warns. Use action to distinguish operations at a merchant. Stored: IDs, keyed fingerprints and amounts, not raw action/merchant. Records persist until the API key is deleted; keyless records have no automatic expiry. $0.002 per call; shared 3 free calls per UTC day without a key (daily private IP namespace), plus 10/day with a free key. Anonymous scopes rotate daily; use a key for a persistent ledger/chain. Send the same Bearer key plus a payment header to keep the same ledger/chain after the allowance runs out.
Annotations the server declares
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
payment_checkPayment check: is this seller safe to pay, checked at the moment the agent pays.Reads text others write1 finding
Why it is in this group
- The name "payment_check" says "payment": a word that usually means money moves, which cannot be taken back.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
Findings
- Warn
The server marks "payment_check" read-only (readOnlyHint: true), and its own name, description or input give the class "irreversible".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":true}; The name "payment_check" says "payment": a word that usually means money moves, which cannot be taken back.
Description, as the server wrote it Payment check: is this seller safe to pay, checked at the moment the agent pays. Compares the 402 price and payee with the seller's published listings when they can be fetched, screens the payee against a local denylist and the bundled OFAC SDN digital-currency address list, flags an over-price versus the intended amount, reuses the duplicate-payment check, and confirms the seller URL returns a valid 402. Returns allow, review or block with reasons and a signed receipt id. Fetches time out in about 2s and fail closed to review. $0.005 per call.
Annotations the server declares
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: true
publish_receiptPublish a signed receipt you own so a public page and a Screwtape Verified badge can link to that one verdict.Reads text others writeCan send data outNo findings
Why it is in this group
- The name "publish_receipt" says "publish": a word for a change that another call does not undo.
What it can reach
- untrusted-inputThe server marks it openWorldHint: true: it reaches things outside the server's own data.
- outboundThe name "publish_receipt" says "publish": it sends or publishes something to other people.
Description, as the server wrote it Publish a signed receipt you own so a public page and a Screwtape Verified badge can link to that one verdict. Call it when you want to show that a specific output received a specific verdict. Do NOT call it to certify an agent in general: the page says this output received this verdict at this time, and it is not a general safety certification of the agent. Nothing is public until this call. Inputs: receipt_id (required), include_content (optional boolean, default false; true stores and shows the task and agent output only if that call was retained), label (optional, at most 80 characters), unpublish (optional boolean; true removes the public page). Returns public_url, badge_svg_url, embed_html, and embed_markdown. Requires the API key that owns the receipt. Cost: free.
Annotations the server declares
- title: Publish one signed receipt as a Screwtape Verified badge
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: true
- openWorldHint: true
Changes things 6Claude Code: ask
action_receiptAction Receipts: Ed25519-signed, hash-chained attestation of an agent-reported action and status; not evidence the action happened or a legal compliance certification.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 Action Receipts: Ed25519-signed, hash-chained attestation of an agent-reported action and status; not evidence the action happened or a legal compliance certification. Per caller and agent chain; concurrent appends serialize with compare-and-set. Raw action/reference are hashed, not stored. event_id per agent is the authoritative idempotency key; a custom HTTP key cannot override it. Same event_id/body returns the same receipt; changed body conflicts. Optional children (1-64 receipt UUIDs) closes a run: the signed children_root digests ordered child receipt_hash values so any added, dropped or swapped step fails verification. GET /v1/receipt/verify?id=<receipt_id> is free and returns the receipt to anyone who knows its unguessable ID; children_root_valid is set when children are present. A receipt proves the record is unchanged since issued, not that a payment was correct or an outside action happened. Or pass receipt=<JSON> to verify a portable receipt. Keep IDs private if your agent IDs/status are sensitive. Records persist until the API key is deleted; keyless records have no automatic expiry. $0.003 per call; shared 3 free calls per UTC day without a key (daily private IP namespace), plus 10/day with a free key. Anonymous scopes rotate daily; use a key for a persistent ledger/chain. Send the same Bearer key plus a payment header to keep the same ledger/chain after the allowance runs out.
Annotations the server declares
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
check_duplicate_invoiceSpend gate for invoices: decides whether an agent may pay this invoice, with a signed receipt as proof.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 Spend gate for invoices: decides whether an agent may pay this invoice, with a signed receipt as proof. Same as judge with one duplicate check: it compares this invoice with the invoices this caller recorded earlier and also judges it. Call it before an agent pays or approves an invoice or payment instruction. Inputs: the judge inputs (task, agent_output as the invoice JSON, evidence, criteria, checks, mode, tags) plus duplicate {vendor_path, invoice_id_path, amount_path, currency_path?, date_path?, payee_account_path?, accept_payee_change?, window_days?, store?, retain_days?}. Returns the judge result plus duplicate.status (unique, duplicate, possible_duplicate, replay or not_checked). On duplicate or possible_duplicate, next_step.action is escalate: do not pay. Fingerprints are kept per caller; only keyed HMAC fingerprints are stored, and only when store is true. Cost: 1 call, the same as judge. First 3 calls per UTC day need no key (daily private IP namespace). Then use Authorization: Bearer <key>, or a per-call payment in params._meta with no key (records are kept per payer; an unverified payment returns the payment requirements again). A free key covers 10 calls per UTC day (get_api_key or https://screwtape.ai/signup/). After the shared 3/day anonymous allowance (3/minute, global 2000/day), tools/call without a key or payment is HTTP 200 with isError true and the payment requirements in structuredContent.
Annotations the server declares
- title: Check an invoice for duplicates before the agent pays
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: true
consensusFalse-consensus check for 2 to 16 answers to the same task, so a group of agents does not approve a spend just because they copied each other.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 read-only (readOnlyHint: true), but an annotation never lowers a class.
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 False-consensus check for 2 to 16 answers to the same task, so a group of agents does not approve a spend just because they copied each other. It scores how the answers agree and flags agreement that looks copied or too thin to trust. Screwtape is not tied to the agent framework or the model that wrote the answers. Call it when several agents, samples or tools answered the same question and you are about to accept the majority. Do NOT call it with a single answer (use judge), with answers to different tasks, or as proof that an answer is correct: high agreement is not evidence. Inputs: task, answers (2 to 16 objects {agent_id?: string, output: string}), optional evidence (a string or an array of strings, used when scoring), retain, and optional call-level agent_id and version tags. The call-level agent_id is not the per-answer agent_id. Returns: consensus_answer, agreement_score, false_consensus_risk (low, medium or high), reasons, clusters, engine, request_id, retention, the tags when sent, and receipt when signing is configured. In that receipt, verdict is false_consensus_risk and score is agreement_score. If the risk is medium or high, verify consensus_answer with judge before using it. next_step.action is pass for low risk and escalate otherwise. Cost: 1 call. First 3 calls per UTC day need no key; then use Authorization: Bearer <key> or pay per call. A free key covers 10 calls per UTC day (get_api_key or https://screwtape.ai/signup/). After the shared 3/day anonymous allowance (3/minute, global 2000/day), tools/call without a key or payment is HTTP 200 with isError true and the payment requirements in structuredContent.
Annotations the server declares
- title: False-consensus check across 2 to 16 agent answers
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: true
get_api_keyCreate a free Screwtape.ai API key (10 calls per UTC day) with no human approval.No findings
Why it is in this group
- The name "get_api_key" says "get": a word for reading.
- The description's first verb is "create": a word for a change.
Description, as the server wrote it Create a free Screwtape.ai API key (10 calls per UTC day) with no human approval. Call it only when no Screwtape key is configured and you can compute SHA-256 with a code tool, because the proof of work must be computed exactly, not guessed. Do NOT call it when a key already exists, to get more quota (it is limited to 5 keys per IP per hour and per UTC day), or without a valid pow_nonce. A missing nonce returns an error that describes the challenge. Inputs: pow_nonce (required string matching ^[A-Za-z0-9_-]{8,64}$; SHA-256 of the UTF-8 string screwtape-signup-v1:<current UTC YYYY-MM-DD>:<nonce> must start with 12 zero bits, which is 3 hex zeros), email (optional), name (optional, at most 80 characters). Returns: api_key (shown once; store it securely and send it as Authorization: Bearer), key metadata and request_id. Only the key's SHA-256 hash is stored. Cost: free. Needs no key and no payment.
Annotations the server declares
- title: Get a free API key (proof of work, no human approval)
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
judgeA pass/fail/retry verdict on ONE agent output, from a service outside your agent, used as the go/no-go before an agent spends money or acts on that output.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 read-only (readOnlyHint: true), but an annotation never lowers a class.
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 A pass/fail/retry verdict on ONE agent output, from a service outside your agent, used as the go/no-go before an agent spends money or acts on that output. Call it before you pay on, return, store or act on a draft answer, especially when you have the evidence the answer should be based on. Do NOT call it to look up facts (it does no web search and only checks against the evidence you pass), to compare several agents' answers (use consensus), or for offline evaluation over a dataset. Inputs: task, agent_output, evidence, criteria, retain, and optional agent_id and version tags. Returns: verdict, score, reasons, engine, next_step, request_id, retention, the tags when sent, and receipt when signing is configured. On pass, use the output. On retry, revise using reasons and call again, at most 2 or 3 times. On fail, stop. Optional checks run first; on failure you get next_step.hint saying what to fix. Branch on next_step.action (pass, retry or escalate). Cost: 1 call. First 3 calls per UTC day need no key; then use Authorization: Bearer <key> or pay per call. A free key covers 10 calls per UTC day (get_api_key or https://screwtape.ai/signup/). After the shared 3/day anonymous allowance (3/minute, global 2000/day), tools/call without a key or payment is HTTP 200 with isError true and the payment requirements in structuredContent.
Annotations the server declares
- title: Judge one agent output (pass, fail or retry)
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: true
try_checkSee what a spend check returns before you have a key.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 See what a spend check returns before you have a key. Runs ONE sample check on a fixed demo invoice: a duplicate match against an invoice paid earlier, required fields, and a sum check that the line items add up to the total. Call it once to learn the response shape. Do NOT call it on your own invoices (it ignores them; use check_duplicate_invoice) or in a loop: it is limited to 5 calls per hour per IP and per MCP session. Inputs: scenario (optional: duplicate, the default, clean, or bad_total). Returns: demo, scenario, invoice, earlier_invoice, duplicate.status (unique or duplicate), checks, next_step.action (pass, retry or escalate), note, next_action and request_id. No receipt and nothing is stored. Cost: free. Needs no key.
Annotations the server declares
- title: Try one sample duplicate-invoice check (free, no key)
- 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 write5
- 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 20, 2026, 18:27 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
- 11
- 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 11 tools that change things (4 irreversible, 7 write). Whether a call also works without credentials was not tested: tools are never called.
What this check looks forMatched text
https://screwtape.ai/mcp: tools/list returned 11 tools to an anonymous client, including payment_check (irreversible), payment_guard_authorize (irreversible), budget_ledger (irreversible), publish_receipt (irreversible), trust_score (write) and 6 more
Posture changes
- Nov 20, 2026
- Tools that change things, listed without credentials: 9 to 11
- Nov 13, 2026
- Tools that change things, listed without credentials: 6 to 9
- Oct 29, 2026
- Tools that change things, listed without credentials: 4 to 6
Findings
Matches of fixed checks, with the text they matched.
- Warnpayment_check
The server marks "payment_check" read-only (readOnlyHint: true), and its own name, description or input give the class "irreversible".
What this check looks forMatched text
annotations: {"readOnlyHint":true,"destructiveHint":false,"idempotentHint":true,"openWorldHint":true}; The name "payment_check" says "payment": a word that usually means money moves, which cannot be taken back.
History
Changes to its tool definitions and to its remote endpoint, newest first.
PostureRemote posture changed. Tools that change things, listed without credentials: 9 -> 11.View diff
- Tools that change things, listed without credentials: 9 to 11
PostureRemote posture changed. Tools that change things, listed without credentials: 6 -> 9.View diff
- Tools that change things, listed without credentials: 6 to 9
Tools1 added.+1 addedView diff
- Added:payment_checkCan't be undone
Tools1 changed.~1 changedView diff
- Changed:action_receiptChanges thingsdescriptioninputSchema
Tools3 added, 1 removed, 4 changed.+3 added−1 removed~4 changedView diff
- Added:action_receiptChanges things
- Added:budget_ledgerCan't be undone
- Added:try_checkChanges things
- Removed:buy_pack
- Changed:check_duplicate_invoiceChanges thingsdescriptioninputSchema
- Changed:consensusChanges thingsdescriptioninputSchema
- Changed:judgeChanges thingsdescriptioninputSchema
- Changed:publish_receiptCan't be undonedescription
Tools1 added, 4 changed.+1 added~4 changedView diff
- Added:check_duplicate_invoiceChanges things
- Changed:buy_packChanges thingstitledescriptioninputSchemaannotations
- Changed:consensusChanges thingsdescriptioninputSchema
- Changed:get_api_keyChanges thingsdescriptioninputSchema
- Changed:judgeChanges thingsdescriptioninputSchema
PostureRemote posture changed. Tools that change things, listed without credentials: 4 -> 6.View diff
- Tools that change things, listed without credentials: 4 to 6
Tools1 changed.~1 changedView diff
- Changed:judgeChanges thingsinputSchema
Install
Pinned to version 1.4.0. Values in angle brackets are yours to fill in.
claude mcp add --transport http screwtape https://screwtape.ai/mcpclaude mcp add --transport http screwtape https://screwtape.ai/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": [],
"ask": [
"mcp__screwtape__budget_ledger",
"mcp__screwtape__action_receipt",
"mcp__screwtape__payment_check",
"mcp__screwtape__judge",
"mcp__screwtape__check_duplicate_invoice",
"mcp__screwtape__consensus",
"mcp__screwtape__get_api_key",
"mcp__screwtape__publish_receipt",
"mcp__screwtape__try_check"
],
"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 screwtape. Review it before you use it.
Capability tags behind 5 rules
- mcp__screwtape__payment_check untrusted-input
- mcp__screwtape__judge untrusted-input
- mcp__screwtape__check_duplicate_invoice untrusted-input
- mcp__screwtape__consensus untrusted-input
- mcp__screwtape__publish_receipt 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/ai.screwtape/screwtape/verify?hash=43ba03b4dfdb1bb8ec037bc9b2ae801de1a60f22022a61e7bfd4f61b4cd9525a" | jq -e '.changed == false'Pinned hash 43ba03b4dfdb, from Nov 10, 2026, 04:18 UTC.
Watch
Get told when its tools change: a new tool, a new description, a new class.
https://radar.anlyon.com/servers/ai.screwtape/screwtape/changes.xmlAn Atom feed. Works in any feed reader or Slack's RSS app.
Badge
[](https://radar.anlyon.com/servers/ai.screwtape/screwtape)Source: MCP Registry. Synced Nov 10, 2026, 04:17 UTC.
Screwtape.ai MCP server: questions and answers
Answered from the scan above, not from the publisher's description.
Is the Screwtape.ai MCP server safe?
- No scan can promise that. What its definitions show: 9 tools, 9 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 Screwtape.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 Screwtape.ai MCP server do things that cannot be undone?
- Yes. 3 tools can send, pay, publish or merge: budget_ledger, payment_check and publish_receipt.
What tools does the Screwtape.ai MCP server have?
- 9 tools: 3 irreversible, 6 write. budget_ledger, payment_check, publish_receipt, action_receipt, judge, check_duplicate_invoice, consensus, get_api_key and 1 more.
Does the Screwtape.ai MCP server need sign-in or OAuth?
- Its remote endpoint listed its tools to an anonymous client, without sign-in.
Does the Screwtape.ai 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): publish_receipt.
Has the Screwtape.ai MCP server changed its tools?
- Yes, 6 times since Anlyon Radar started tracking it, most recently on Nov 10, 2026. Every change is on this page and in its Atom feed.