Agent Audit MCP
Tamper-evident, hash-chained audit log for agent actions, exposed over MCP
- Pricing
- Paid
- Platforms
- CLI, self-hosted, API
- Version
- 1.0.0
- Listing updated
- today
Similar MCP servers
About Agent Audit MCP
Agent Audit MCP is a paid MCP server for security. It is available on CLI, self-hosted and API.
The README describes a local log where every agent event carries a SHA-256 hash that includes the previous event's hash, so editing or removing an entry breaks every later link. It covers sixteen event types spanning tasks, payments, disputes, identity checks and messages, each tagged with one of five severity levels for triage.
An MCP client can write events, fetch one event or an agent's full history, search by agent, type, severity or date range, and run a verification pass that reports PASS or FAIL for each link between two events. A summary tool gives totals by type and severity plus the most active agents.
The README says data is kept in a local JSON file and that the server makes no network requests. It is aimed at developers and operators who want a provable record of what their agents did, for compliance review or dispute handling.
Tools this server provides (6)
Taken from the project documentation; the running server may differ.
- audit_log
- Log an event and append it to the hash chain
- audit_get_event
- Retrieve full event details by ID
- audit_search
- Filter events by agent, type, severity, or date range
- audit_get_agent_history
- All events for one agent, newest first
- audit_verify_chain
- Check SHA-256 chain integrity between two events; returns per-event PASS/FAIL
- audit_stats
- Summary of total events, breakdown by type and severity, top 5 active agents
What it does
- Hash-chained event log
- Each event's SHA-256 hash covers its fields and the previous event's hash, so tampering is detectable.
- Per-link chain verification
- Recomputes hashes between two events and reports pass or fail for each event.
- Sixteen agent event types
- Covers task, payment, dispute, identity, capability and messaging events.
- Severity-based triage
- Events are tagged info, notice, warning, error or critical for compliance review.
- Flexible search
- Filter by agent, event type, severity or date range, with a result cap.
- Local data storage
- The README says data stays in a local file and no network requests are made.
Good for
- Keep a tamper-evident record of agent payments and task outcomes
- Investigate disputes by reviewing an agent's full ordered history
- Check that a stored audit trail has not been altered
- Triage agent errors and critical events by severity
- Review which agents are most active across a deployment
Pricing
Paid, from $19 a month
Pro plan at $19/month; unlimited audit events and full chain verification.
Agent Audit MCP FAQ
Where is audit data stored?
The README says data is stored locally at ~/.agentaudit/chain.json and that no data leaves the machine.
Which clients can connect to it?
The README gives setup instructions for Claude Desktop and Cursor using a stdio command pointing at server.py.
What does verification check?
audit_verify_chain recomputes each hash from scratch between two event IDs and returns a per-event PASS or FAIL result.
What does it cost?
The README lists a Pro plan at $19 per month with unlimited events and full chain verification, subscribed via Stripe. The license is proprietary.
Is Agent Audit MCP free?
No. Agent Audit MCP is a paid product. Paid plans start at $19 a month.
What are the best alternatives to Agent Audit MCP?
The closest alternatives to Agent Audit MCP are AgentProof, Web3 Systemic Agent, Trust Layer, Tracekit MCP gateway and Sovereign Agent Logbook.
Where this comes from
- Description: written from the project's README and its MCP Registry entry.