Runvouch MCP server
{
"mcpServers": {
"runvouch": {
"type": "http",
"url": "https://api.runvouch.com/mcp"
}
}
}{
"mcpServers": {
"runvouch": {
"type": "http",
"url": "https://api.runvouch.com/mcp"
}
}
}{
"mcpServers": {
"runvouch": {
"type": "http",
"url": "https://api.runvouch.com/mcp"
}
}
}{
"servers": {
"runvouch": {
"type": "http",
"url": "https://api.runvouch.com/mcp"
}
}
}{
"mcpServers": {
"runvouch": {
"type": "http",
"url": "https://api.runvouch.com/mcp"
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsonversion 0.3.4
What it does
RunVouch is a monitoring server for unattended AI agents that detects failures, budget overruns, and suspicious behavior, then alerts via email, Slack, Telegram, or webhooks. It generates tamper-evident SHA256 records of each run, chained daily and timestamped in Bitcoin. Use it to verify that scheduled AI tasks actually completed as intended, not just that they ran.
- Detect when a scheduled agent misses its run, fails, or loops repeatedly without stopping.
- Verify that an agent produced required evidence (a file, HTTP response, or assertion) before reporting success.
- Alert when a run exceeds a cost cap or shows unusual duration or output size compared to baseline behavior.
Tools this server exposes
| Tool | Does |
|---|---|
| runvouch_status | Query the status of registered agents and their recent runs. |
| runvouch_alerts | Fetch active or historical alerts for agent runs. |
| runvouch_ack | Acknowledge an alert to silence it. |
| runvouch_runs | List runs for a specific agent. |
| runvouch_run_start | Record the start of an agent run with metadata. |
| runvouch_run_end | Record the end of an agent run with exit status and evidence. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.