What Did My Agent Do?

Audit an MCP tool call

MCP server logs show what you saw. See what another party could independently verify about a tool call, and how to capture it.

Analyze an MCP tool call →

Compare internal MCP observability with cross-party evidence.
Learn where to attach a signed record to a tool-call response.
Copy a minimal snippet to sign records at your service boundary.

Example input

[14:23:01] mcp.tool_call: check_inventory
[14:23:02] upstream: GET https://api.vendor.com/v2/inventory/sku-4419 -> 200
[14:23:02] policy_eval: allow (tools.check_inventory.default)

What the checker shows

Observable: tool call, upstream status, policy evaluation. Independently verifiable: nothing. Missing: signature, issuer, policy hash.

What is missing for cross-party proof

Evidence	Status
Tool call happened	observable, not verifiable
Policy in force at call time	missing (add a policy hash)
Signed record on the result	missing

Close the gap

res._meta["org.peacprotocol/receipt_jws"] = jws; // attach to the MCP tool result

Analyze an MCP tool call →

Show your records are verifiable

Issuing signed records already? Add a badge to your README:

[![agent records: verifiable](https://whatdidmyagentdo.com/badge.svg)](https://whatdidmyagentdo.com)

PEAC interaction records can be carried in MCP _meta alongside tool results.

Open the verifier · Full checker · How it works · PEAC standard