I'm building AgentTrail: observability and guardrails for AI coding agents.
- AgentTrail Guard checks every command and file change from Claude Code, Cursor and Codex CLI against 74 guardrails, locally, before it runs. Open source, Apache 2.0, no account.
- AgentTrail OS traces every agent session and turns repeat failures into rules you can backtest on real history before enforcing them.
npm i -g @agenttrail/guard
agenttrail-guard init --agent claude # or cursor, codexI've also run TechSuite, a fintech engineering studio, since 2016. We provide fractional CTOs and senior engineers for founders shipping financial products. AgentTrail started there, from running coding agents on real client work.
Before AgentTrail: Head of Engineering at Namu, founded the Carrier Integrations team at Canopy Connect, and wrote SneakerBot (190+ forks).
Ever had an agent run git reset --hard on uncommitted work, or --force-reset a dev database? Book an AgentTrail intro and I'll show you how to stop the next one before it runs.
