
Countersign
io.github.countersign-networkPublished by GitHub account countersign-networkv0.2.2Updated Oct 24, 2026
Kill switch + spend guard for AI agents that spend money, across every wallet vendor at once.
Overview
MCP tools for a Countersign control plane that lets an assistant pre-check agent spending, freeze all agent wallets at once, and review a hash-chained audit…
- What you can do
- Ask whether a proposed spend is allowed before touching a wallet, getting allow, deny or needs_approval.
- Freeze or unfreeze every governed agent across all wallet backends at once.
- Apply one unified policy of caps, allow/deny lists and approval thresholds.
- List governed agents and check control-plane health.
- Work the human-approval queue: list, approve and deny pending requests.
- Read and re-verify the append-only, hash-chained audit ledger.
- When to use it
- Useful when an assistant drives agents that move money and you want a single pre-flight spend check, a one-call kill switch and an auditable record across several wallet or card backends. It is a governance layer, not a wallet itself.
- Requirements
- Runs locally via npx from the npm package @countersign/mcp over stdio. Needs a running Countersign Core: set COUNTERSIGN_URL to its base URL, and COUNTERSIGN_API_KEY when that Core has auth enabled. The hosted Core is described as paused and returning 503, so a self-run Core is expected. Desktop MCP clients only.
Installation
In SourceWeft
- Open Countersign in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
@countersign/mcp
[npm version] [MCP] [License: Apache-2.0] [Spending governed by Countersign]
The cross-vendor kill switch + spend guard for AI agents — as MCP tools, inside any MCP client.
Countersign is a neutral control plane that holds one policy, one freeze, and one tamper-evident audit ledger across every agent-wallet backend at once. This package exposes it as Model Context Protocol tools, so an agent (or you, from chat in Claude / Cursor) can ask "may I spend this?" before acting — and freeze everything in under a second.
Tools
Setup
It governs a running Countersign Core. The hosted Core at app.countersign.network is paused since
2026-09-18 and answers 503, so point this at a Core you run yourself. Set:
COUNTERSIGN_URL— your Core, e.g.https://app.countersign.networkCOUNTERSIGN_API_KEY— required when the Core has auth enabled
Claude Desktop / Claude Code / Cursor
Add to your MCP config (claude_desktop_config.json, or claude mcp add, or Cursor's mcp.json):
That's it — your agent now has the spend guard and the kill switch.
Show it (the "powered by" badge)
If Countersign guards your agent, say so — it tells users (and other agents) that spending is governed, and points the next operator at the same kill switch. Drop this in your README:
Listing in MCP registries (distribution)
Countersign spreads by being easy to discover wherever agent builders look for tools. The metadata below is ready to submit to the common registries — copy it as-is.
- Official MCP registry (
registry.modelcontextprotocol.io/ themcp-registryrepo) — entry: - mcp.so — submit
@countersign/mcp; category Finance / Security / Agents. - Smithery —
npx-launched server; envCOUNTERSIGN_URL(defaulthttps://app.countersign.network) + optionalCOUNTERSIGN_API_KEY. - awesome-mcp-servers lists — one-liner:
Countersign — cross-vendor spend guard + sub-second kill switch + tamper-evident ledger for AI agents that move money (Coinbase, Turnkey, Openfort, Visa card). Testnet.
Standard listing fields:
Nothing here is auto-submitted — listings are made deliberately by a maintainer.
Apache-2.0. Countersign holds policy/freeze/ledger; it never takes custody of funds. Testnet only.
Source: packages/mcp/README.md at commit 534840a
Tools
0Version history
1- v0.2.2LatestSep 16, 2026
Score
40% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts
Declare a license in the repository or the registry entry.