All servers / Payments / spendctl MCP server
spendctl MCP server
MCP server exposing spendctl to agents: pay, await_approval, withdraw_approval, check_budget, list_receipts. Holds no keys.
Tested by ConnectATDScore 82 out of 100. Tested 2026-11-19 on package version 0.10.0. Package: @spendctl/mcp-server
Score breakdown
| Runs | 18 / 20 | |
|---|---|---|
| Tool descriptions | 20 / 20 | |
| Package safety | 16 / 20 | |
| Maintenance | 20 / 20 | |
| Adoption | 8 / 20 |
What we saw
- Starts and answers the MCP handshake: yes (3.6 seconds including download)
- Tools offered: 5
- Last published: 2026-11-14
- Weekly downloads (npm): 665
- License: SEE LICENSE IN LICENSE · Direct dependencies: 3
Want this server set up for your team?
ConnectATD builds AI automation and wires up MCP servers like this one. Tell us what you are trying to do.
Talk to ConnectATDTools and how they are described
Agents choose tools from these descriptions. Under each name you can see how many of its parameters are explained.
pay7 of 7 parameters describedFetches an HTTP(S) URL that may require payment (HTTP 402: x402 or MPP) through the local spendctl sidecar, which applies the human's budget, allowlist and approval policy, pays if allowed and returns the response. Use it for any URL that might charge; a free URL behaves like a normal fetch. Use dryRun to learn the decision without paying, and the same idempotencyKey when retrying so nothing is paid twice. Returns a one-line summary (HTTP status, amount paid, decision, receipt id, and the Idempotency-Key or X-Request-Id you sent, if any) followed by the response body (clipped at 20,000 characters), or the refusal with its reason, or an approval id when a human must approve first. The decision is the sidecar's: a refusal or a hold cannot be bypassed through this tool, and the keys never leave the sidecar. Requires a running sidecar (`spendctl start`). Example: url https://api.example.com/v1/weather?city=Kyoto
await_approval2 of 2 parameters describedWaits for a human to decide a payment that pay reported as held (Approval id: apr_…), without paying again. Long-polls the sidecar for up to timeoutSeconds (default 20, at most 25; keep it under your client's tool timeout) and returns the approval's status: pending (call again), approved (repeat pay with approvalId to complete the payment), denied, expired or withdrawn (this agent took it back with withdraw_approval). Only this agent's own approvals are visible and no admin token is involved. Read-only: it never pays and never decides.
withdraw_approval2 of 2 parameters describedTakes back a payment that pay reported as held (Approval id: apr_…) when the task that needed it was cancelled or superseded, so a late human approval cannot pay for it. Nothing is paid: the approval and its ledger entry close as withdrawn (distinct from a human's denial), a later `spendctl approve` on that id is refused, and await_approval on it answers withdrawn at once. Only this agent's own pending approvals can be withdrawn, with the agent's identity and no admin token. If the payment is needed after all, a new pay call opens a fresh approval. Withdrawing an id that was already decided reports that decision as an error.
check_budgetno parametersShows this agent's spending limits, what has been spent in each window and what is left. Use it before a large or repeated spend, 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.
list_receipts1 of 1 parameter describedLists this agent's most recent payment attempts, newest first: paid (amount and transaction hash), refused (reason codes) and waiting for approval. Use it to report what was paid or to find the approval id a human needs. Returns up to `limit` lines, one per attempt, with time, receipt id, method, host and path, amount and status, plus the Idempotency-Key and X-Request-Id the request carried, if any, or "No payments yet." Read-only.
Tested in a sandbox with an empty environment: no API keys, a throwaway folder, nothing from a real account. A passing test shows the server runs and responds. It is not a security audit.