Dr. Claw: an open-source agent workbench covering the whole research pipeline
aigclink · x · 2026-09-14
Dr. Claw is an open-source (AGPL) research agent workbench positioned as a Claude Science alternative: swap in Claude Code / Gemini / Codex as the base model, with human-in-the-loop control, rollback and full audit trails.
Pipeline coverage:
- Review: literature survey reports with arXiv / Semantic Scholar citations
- Idea selection: SCAMPER and SWOT brainstorming scored by multiple personas
- Experiments: plan → implement → judge loop producing experiment code, statistical analysis and paper-grade figures
- Paper: IEEE/ACM manuscripts with LaTeX formulas, citation audit and Overleaf sync
- Outreach: academic slides, TTS narration, demo video and project page
Placed in the 'AI scientist' race: Faraday trains a closed 27B 'research taste' model, OpenRSI trains weights plus long-horizon search, while Dr. Claw bets the bottleneck is workflow, not the model. Useful for universities, corporate R&D standardization, and builders can fork it.
More from coding & agent
- Could multi-agent AI breed groupthink and block an 'Einstein' agent from emerging? — Genzinvestor16180339 · 2026-09-14
- Satya Nadella touts Microsoft Foundry as the platform to make enterprise agents governable — satyanadella · 2026-09-14
- Free Aura components MCP: 2,495 UI components and 725 design systems for your coding agent — Next_Token_AI · 2026-09-14
- Yacine calls out LLMs: why isn't high-quality code the default output? — yacineMTB · 2026-09-14
- Seeking Open-Source Models to Replace LiveKit Adaptive in a Local Voice Agent — Johnstoncm · 2026-09-14
- OpenClaw scheduled agent task fails on approval policy while the same command works interactively — Grimmoner · 2026-09-14