TracePact – Catch tool-call regressions in AI agents before prod
VCR cassettes for agent tool sequences—catches prompt regressions before deploy.

Runs actual code in sandboxes instead of static analysis or URL-clicking agents.
Engineering teams, QA leads, senior developers
Cursor bugbot · Greptile · Coderabbit
It's code review that actually runs your code. The result is that it finds more bugs with a smaller false positive rate. You get screenshots, videos, and run logs. I built Ito because manual verification sucks and I thought I could automate it. I've been a professional developer for 15 years and now more than ever, I'm just spending time making sure things work.
I looked into all of the other agentic testing tools that click around a url you give them and frankly, they sucked. They just didn't work. I figured to be able to actually testing things you need to run the full environment so you can seed data, bypass gates, mock external dependencies. They also can't get runtime evidence about changes that aren't frontend. Ito does.
Under the hood, Ito is creating a devcontainer for your application. All of your services are wired up and then a team of agents goes and tests your PR. We rolled out our own sandboxes and our own agentic coding loops.
I built Ito to be another useful signal in development. It shouldn't annoy you when it's not perfect and it saves you a bunch of time testing.
A few open source examples:
- n8n https://app.ito.ai/share/bb7d73aa-fd08-482d-9938-87938e2a232...
- lobehub https://app.ito.ai/share/1d76f15a-9998-4154-85b4-b177924cba8...
- excalidraw https://app.ito.ai/share/d1cb1475-fbe5-4c71-901b-409ba2aa6d6...
VCR cassettes for agent tool sequences—catches prompt regressions before deploy.
Personal framework for one AI assistant — clever but too narrow to generalize.
Behavioral safety testing reveals 45 regressions static analysis misses—guardrails provided.
Useful Claude Code skill, but it's config for an existing tool, not a product.
Docker sandbox records every agent action so you can replay suspicious behavior.
MicroVM sandboxes with secret proxying beats Git worktrees for agent isolation.