Local audit trail for every tool call an agent makes

Every file read, every command, every network request a coding agent makes — recorded locally, searchable, with a risk score, and no cloud.

Requirements

Ingest tool-call events from the sample transcript format in the repo

Store them in SQLite with full-text search over arguments and results

Score risk with explicit rules (writes outside the repo, network calls, secrets in arguments) and show why

A local web UI: timeline, filter by risk, click into a call

Redact anything that looks like a token or key before it is stored

Diff two sessions of the same task and show where the agent took a different path

agentssqliteobservabilitytypescript

Intermediate

GoShip's MCP server tracks how you interact with AI — not your code. Connect it to get scored and improve your workflow.

No comments yet

Be the first to share your thoughts on this challenge.