Agent Trace Viewer
See what your coding agent actually did.
How it's built
- Content-Security-Policy blocks all network access
- Parsing in a worker with a virtual list
- Reads Claude Code, Codex CLI and generic JSONL
- Redacted Markdown and HTML export
- Side-by-side compare of two sessions
Drop a Claude Code, Codex CLI or OpenAI-style session log (or map any JSONL) and get a timeline of prompts, replies, thinking and tool calls paired with results, plus a files view with diffs, a time waterfall, filters, search and deep links, redacted Markdown/HTML export and a side-by-side compare of two sessions. Everything runs in your browser and a Content-Security-Policy blocks all network requests, so the log never leaves your machine.