An open-source AI lab shipping tools, agents, and experiments across the AI development stack.
Tools that change how your coding agent works — and show you what it is actually doing.
- autoharness (801 stars) — A self-learning skill layer for Claude Code: distills skills from your real sessions, updates them as you work, and prunes the ones that go unused.
- cost-xray (367 stars) — See what Claude Code and Codex actually send to the API — and what each part costs.
Skills that put the agent to work on everything around the code — design, research, distribution.
- design-harness (182 stars) — Feed your agent papers and half-formed ideas; it links them into a system design you can defend. An Agent Skill for Claude Code and any SKILL.md-compatible agent.
- auto-gtm (107 stars) — A Claude Code / Codex plugin that drafts your X and Reddit posts from your PRs and the day's hot threads — drafts only, never posts for you.
- paper-radar (1 stars) — Finds the AI papers 28 labs put on arXiv over any date range, and separates the ones a company led from the ones it only appears on.
- influencer-discovery (0 stars) — Finds creators who bring their own audience — followers, readers, subscribers — across 15 channels, and appends their public contact info to a Google Sheet. An Agent Skill for Claude Code.
- phi-boundary-gate (56 stars) — PHI boundary reports for healthcare and insurance AI workflows — across prompts, RAG, tools, memory, and logs.
More in the works — built in the open, shipped fast.
