I build AI agent systems, browser automation tools, and local-first developer workflows.

Local flight recorder for AI browser agents.

BrowserTrace helps Browser Use, Stagehand, Skyvern, Playwright + LLM, and custom computer-use builders debug failed browser-agent runs with local step timelines.

• Records screenshots, URLs, actions, model input/output, status, and errors.
• Opens failed runs in a local web UI.
• Exports standalone HTML traces.
• Supports public-safe exports that omit prompts, model I/O, screenshots, and URLs.
• Supports Browser Use callback and run-hook tracing.
• MIT licensed and local-first.

Repo: https://github.com/aaronlab/browsertrace

Live demo: https://aaronlab.github.io/browsertrace/

Try locally from PyPI:

uvx --from "browsertrace[ui]" browsertrace doctor
uvx --from "browsertrace[ui]" browsertrace demo

Persistent install:

pip install "browsertrace[ui]"
browsertrace doctor
browsertrace demo

Computer-use guide: https://aaronlab.github.io/browsertrace/computer-use-agent-debugging.html

Runnable examples: https://github.com/aaronlab/browsertrace/tree/main/examples

Roadmap: https://github.com/aaronlab/browsertrace/blob/main/ROADMAP.md

Launch feedback: aaronlab/browsertrace#3

Good first issue: https://github.com/aaronlab/browsertrace/labels/good%20first%20issue

First PR Recipe: https://github.com/aaronlab/browsertrace/blob/main/CONTRIBUTING.md#first-pr-recipe keeps the first contribution small and reviewable.

Public-safe demo export: https://github.com/aaronlab/browsertrace/releases/download/v0.1.17/browsertrace-demo-public.html

For profile-reader follow-up, local first-run issues, CI failures, or AI/coding-agent troubleshooting replies, ask for debugging/workflow details plus JSON CLI diagnostics when safe to share:

browsertrace doctor --json
browsertrace list --status failed --json
browsertrace show <run_id> --json

• AI browser-agent debugging
• Browser automation and computer-use agents
• LLM observability for local workflows
• Agent evaluation and tool reliability

If you build browser agents, the most useful BrowserTrace feedback is:

• Which framework do you use?
• What context is missing when a run fails?
• Are local HTML exports enough, or do you need hosted share links?
• Which adapter should be improved first?

Launch discussion: aaronlab/browsertrace#6