spendctl MCP server
dev.spendctl/spendctlspendctl.devstreamable-httpv0.1.0Scanned
Spend policy for AI agents that pay (x402, MPP). Login-free sandbox, real sidecar, no real funds.
8 tools: 2 do something that can't be undone, 2 change things and 4 only read.
- Reads private data · 0
- Reads text others write · 0
- Can send data out · 1
- 1 finding
Tools listed without sign-in. Remote posture
With Anlyon
2 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 spendctl MCP server can do
The spendctl MCP server is published by spendctl.dev, a domain the MCP Registry verified. It has 8 tools. It can do things that cannot be undone, such as sending, paying or publishing (sandbox_pay and withdraw_approval) and create or change data (2 tools: sandbox_create and approve), but cannot delete data. 4 tools only read.
1 tool can send data out of your system (sandbox_pay).
The fixed checks matched 1 finding (tool-shadowing). Each one comes with the text it matched.
Its remote endpoint lists its tools without sign-in, and its tools changed once since Anlyon Radar first saw them.
Last scanned Nov 9, 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.
8 tools
Can't be undone 2Claude Code: ask
sandbox_payFetches one of the sandbox's paid endpoints (HTTP 402, x402) through the spendctl sidecar, which applies the sandbox agent's budget, allowlist and approval policy, pays with the th…Can send data outNo findings
Why it is in this group
- The name "sandbox_pay" says "pay": a word that usually means money moves, which cannot be taken back.
What it can reach
- outboundThe input takes a URL (input.url): a request to a URL can carry data to whoever runs it.
Description, as the server wrote it Fetches one of the sandbox's paid endpoints (HTTP 402, x402) through the spendctl sidecar, which applies the sandbox agent's budget, allowlist and approval policy, pays with the throwaway wallet if allowed, and returns the response. Use it for every paid call in the sandbox: with dryRun to see the decision without paying, with approvalId to complete a payment that was held and then approved. The endpoints are https://weather.sandbox.spendctl.dev/weather?city=<name> ($0.01) and https://search.sandbox.spendctl.dev/search?q=<query> ($0.02), as listed by sandbox_create and documented at https://spendctl.dev/docs/sandbox/; any other host is refused. Returns a one-line summary (HTTP status, amount paid, decision, receipt id) followed by the response body, or the refusal with its reason, or an approval id when the payment is waiting for a human. Only sandbox hosts exist and no real funds are involved; the request is always a GET and responses are clipped at 20,000 characters. Example: url https://weather.sandbox.spendctl.dev/weather?city=Kyoto
Annotations the server declares
- title: Fetch a paid sandbox URL
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
withdraw_approvalTakes back a payment that sandbox_pay reported as held, as a real agent does when the task that needed the payment was cancelled or superseded: nothing is paid, the approval closes…No findings
Why it is in this group
- The name "withdraw_approval" says "withdraw": a word that usually means money moves, which cannot be taken back.
Description, as the server wrote it Takes back a payment that sandbox_pay reported as held, as a real agent does when the task that needed the payment was cancelled or superseded: nothing is paid, the approval closes as withdrawn (distinct from deny), a later approve on that id is refused and await_approval answers withdrawn at once. Use it with an approval id from this sandbox; a new sandbox_pay call opens a fresh approval if the payment is wanted after all. Withdrawing an id that was already decided reports that decision as an error.
Annotations the server declares
- title: Withdraw a held sandbox payment
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
Changes things 2Claude Code: ask
approveApproves or denies a payment that sandbox_pay reported as waiting for approval, as the human approver would with `spendctl approve`.No findings
Why it is in this group
- The name "approve" says "approve": a word for a change.
Description, as the server wrote it Approves or denies a payment that sandbox_pay reported as waiting for approval, as the human approver would with `spendctl approve`. Use it only with an approval id from this sandbox; in the real product a human decides, in the sandbox the model plays that role. Returns the approval id and its new status (approved or denied). An approved payment completes when sandbox_pay is called again with the same approvalId; a denied one stays refused. Deciding the same approval twice keeps the first decision.
Annotations the server declares
- title: Approve or deny a held sandbox payment
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
sandbox_createCreates an isolated spendctl sandbox and returns its sandbox_id.1 finding
Why it is in this group
- The name "sandbox_create" says "create": a word for a change.
Findings
- Warn
The description contains "every other tool", wording about other tools or servers.
What this check looks forMatched text
description: "...returns its sandbox_id. Call it first: every other tool needs the sandbox_id, and one sandbox s..."
Description, as the server wrote it Creates an isolated spendctl sandbox and returns its sandbox_id. Call it first: every other tool needs the sandbox_id, and one sandbox serves a whole conversation (create another only to try a different preset). The sandbox is a real spendctl sidecar with a throwaway wallet, a mock x402 facilitator and two mock paid APIs; no real money moves and nothing outside the sandbox is reachable. Returns the sandbox_id, the agent's policy (budgets, allowed hosts, approval rule) and the paid endpoints with their prices. Limits: 5 sandboxes per minute per client address (claude.ai users share addresses and get a larger allowance); a sandbox expires after 2 hours without calls (24 hours at most). Example: preset "strict" to see refusals, "approvals" to see a payment held for a human.
Annotations the server declares
- title: Create a spendctl sandbox
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
Only reads 4Show 4 read-only tools
await_approvalWaits for the decision on a payment that sandbox_pay reported as held, without paying again: long-polls the sandbox sidecar for up to timeoutSeconds (default 10, at most 20) and re…No findings
Why it is in this group
- The description's first verb is "wait": a word for reading.
Description, as the server wrote it Waits for the decision on a payment that sandbox_pay reported as held, without paying again: long-polls the sandbox sidecar for up to timeoutSeconds (default 10, at most 20) and returns pending, approved, denied, expired or withdrawn, with the ledger entry and the decision deadline. This is what an agent does in the real product while a human decides; in the sandbox the model is also the approver (approve), so call approve from another turn or first. After approved, repeat sandbox_pay with approvalId to complete the payment. Read-only; never pays and never decides.
Annotations the server declares
- title: Wait for a held sandbox payment's decision
- readOnlyHint: true
- openWorldHint: false
check_budgetShows the sandbox agent's spending limits, what has been spent in each window and what is left.No findings
Why it is in this group
- The description's first verb is "show": a word for reading.
Description, as the server wrote it Shows the sandbox agent's spending limits, what has been spent in each window and what is left. Use it before a paid call when the remaining budget matters, or after a refusal to see which limit was hit. Returns the agent name and mode, the per-request limit, the approval threshold, what happens when a budget is exceeded, the allowed hosts and one line per window (hourly, daily: spent, limit, remaining). Read-only; changes nothing.
Annotations the server declares
- title: Check remaining budget
- readOnlyHint: true
- openWorldHint: false
get_policyReturns the sandbox's spendctl.yaml as text: the preset name and the agent's budgets, allowed hosts and approval rules.No findings
Why it is in this group
- The name "get_policy" says "get": a word for reading.
Description, as the server wrote it Returns the sandbox's spendctl.yaml as text: the preset name and the agent's budgets, allowed hosts and approval rules. Use it to explain why a payment was refused or held, or to show what a spendctl policy file looks like. Read-only; the sandbox policy cannot be changed, create a sandbox with another preset instead.
Annotations the server declares
- title: Show the sandbox policy
- readOnlyHint: true
- openWorldHint: false
list_receiptsLists the sandbox agent's most recent payment attempts, newest first: paid (amount and mock transaction hash), refused (reason codes) and waiting for approval.No findings
Why it is in this group
- The name "list_receipts" says "list": a word for reading.
Description, as the server wrote it Lists the sandbox agent's most recent payment attempts, newest first: paid (amount and mock transaction hash), refused (reason codes) and waiting for approval. Use it to report what happened so far or to find a receipt id; after a hold, the approval id is in the sandbox_pay result. Returns up to `limit` lines, one per attempt, with time, receipt id, method, URL, amount and status, or "No payments yet." Read-only.
Annotations the server declares
- title: List receipts
- readOnlyHint: 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 untrusted-input.
- Reads private data0
- Reads files, mail, messages, databases, secrets, calendars or repositories.
- Reads text others write0
- 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 17, 2026, 15:23 UTC, unchanged since Nov 3, 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
- 4
- 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 4 tools that change things (2 irreversible, 2 write). Whether a call also works without credentials was not tested: tools are never called.
What this check looks forMatched text
https://mcp.spendctl.dev/mcp: tools/list returned 8 tools to an anonymous client, including sandbox_pay (irreversible), withdraw_approval (irreversible), sandbox_create (write), approve (write)
Posture changes
- Nov 3, 2026
- Tools that change things, listed without credentials: 3 to 4
Findings
Matches of fixed checks, with the text they matched.
- Warnsandbox_create
The description contains "every other tool", wording about other tools or servers.
What this check looks forMatched text
description: "...returns its sandbox_id. Call it first: every other tool needs the sandbox_id, and one sandbox s..."
History
Changes to its tool definitions and to its remote endpoint, newest first.
Tools2 added, 6 changed.+2 added~6 changedView diff
- Added:await_approvalOnly reads
- Added:withdraw_approvalCan't be undone
- Changed:approveChanges thingsdescriptioninputSchemaannotations
- Changed:check_budgetOnly readsdescriptioninputSchemaannotations
- Changed:get_policyOnly readsdescriptioninputSchemaannotations
- Changed:list_receiptsOnly readsdescriptioninputSchemaannotations
- Changed:sandbox_createChanges thingsdescriptioninputSchemaannotations
- Changed:sandbox_payCan't be undonedescriptioninputSchemaannotations
PostureRemote posture changed. Tools that change things, listed without credentials: 3 -> 4.View diff
- Tools that change things, listed without credentials: 3 to 4
Install
Pinned to version 0.1.0. Values in angle brackets are yours to fill in.
claude mcp add --transport http spendctl https://mcp.spendctl.dev/mcpclaude mcp add --transport http spendctl https://mcp.spendctl.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__spendctl__await_approval",
"mcp__spendctl__check_budget",
"mcp__spendctl__list_receipts",
"mcp__spendctl__get_policy"
],
"ask": [
"mcp__spendctl__sandbox_create",
"mcp__spendctl__sandbox_pay",
"mcp__spendctl__withdraw_approval",
"mcp__spendctl__approve"
],
"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 spendctl. Review it before you use it.
Capability tags behind 1 rule
- mcp__spendctl__sandbox_pay 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.spendctl/spendctl/verify?hash=ab3438474636f5cabd3e2270837e7fae0af1eb0c4e82e47bfc667c88a5f4dc43" | jq -e '.changed == false'Pinned hash ab3438474636, from Nov 9, 2026, 09:17 UTC.
Watch
Get told when its tools change: a new tool, a new description, a new class.
https://radar.anlyon.com/servers/dev.spendctl/spendctl/changes.xmlAn Atom feed. Works in any feed reader or Slack's RSS app.
Badge
[](https://radar.anlyon.com/servers/dev.spendctl/spendctl)Source: MCP Registry. Synced Oct 26, 2026, 09:17 UTC.
spendctl MCP server: questions and answers
Answered from the scan above, not from the publisher's description.
Is the spendctl MCP server safe?
- No scan can promise that. What its definitions show: 8 tools, 4 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 spendctl MCP server delete data?
- No tool is classed as destructive: going by their definitions, none of them deletes, cancels or revokes anything.
Can the spendctl MCP server do things that cannot be undone?
- Yes. 2 tools can send, pay, publish or merge: sandbox_pay and withdraw_approval.
What tools does the spendctl MCP server have?
- 8 tools: 2 irreversible, 2 write, 4 read. sandbox_pay, withdraw_approval, sandbox_create, approve, await_approval, check_budget, list_receipts and get_policy.
Does the spendctl MCP server need sign-in or OAuth?
- Its remote endpoint listed its tools to an anonymous client, without sign-in.
Does the spendctl 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): sandbox_pay.
Has the spendctl MCP server changed its tools?
- Yes, once since Anlyon Radar started tracking it, most recently on Nov 9, 2026. Every change is on this page and in its Atom feed.