cot.run - Self-hosted Observability for AI Coding Agents

Black Hat Arsenal demo of cot.run, a self-hosted observability platform for AI coding agents.

The demo showed how teams can trace sessions from Claude Code, Cursor, and Codex through local hooks, then review timelines, search traces, and understand usage from a dashboard that runs on their own machine.

The session focused on practical visibility for agentic development workflows:

• Capturing local coding-agent activity without sending traces to a third-party service
• Reviewing prompt, tool, and execution timelines across sessions
• Debugging agent behavior when a coding task goes sideways
• Giving security and engineering teams better context around AI-assisted code changes along with insights into usage, cost and security risk patterns.

This was one of the most random demos I’ve given, I took place of someone who had to drop out of the event. I was able to get the demo running in about 5 minutes and presented around 3-4 times with audience always present for the duration of the demo and many questions asked.

Had no presentation prepared, my entire demo was -

• cot.run Website (as the intro of the platform)
• Demo of the platform
• Personal LinkedIn profile as my Introduction Slide
• GitHub Project for feedback and PRs
• Stickers as takeaways

All of this was so so fun!!!