# auditable MCP server

> Provides a library of auditing methods to monitor, detect failures, and localize mistakes in AI agent runs using configuration, online, and post-run analysis.

- Repository: https://github.com/yzhao062/auditable
- Page: https://hostedmcp.org/servers/yzhao062/auditable
- Transport: —
- Stars: 31
- License: Apache-2.0

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `GRADE_DIR` | no | no |  |

## FAQ

### What can auditable audit?

It audits agent configurations before a run, monitors agent behavior while it is running, and detects failures or localizes faults after a run is finished.

### How are the auditing methods organized?

Methods are grouped into four families, with each family sharing a single interface to ensure consistency.

### Can it detect if an agent is over-privileged?

Yes, the configuration audit methods can flag risky permission levels and unrequested sensitive access before a run starts.
