Research workflows are scattered across chats and folders—Dr. Claw wants to bring them into an open-source workbench
Dan Kornas introduces Dr. Claw, an open-source research workbench. It breaks research ideas down into task lists and puts tasks, papers, outputs, and Git operations in a single interface. It is useful for researchers who need to manage state, but may feel heavy for people who just want to chat.
The trouble with research work usually isn't asking AI a question; it's what comes after: literature, tasks, experiments, drafts, reviewer responses, and Git commits are scattered across chat logs, folders, and terminals. Dr. Claw, introduced by Dan Kornas in a tweet, aims to bring these into a single workbench.
It positions itself as an AI research workbench for researchers and builders, covering literature reviews, experiments, paper drafts, and presentation materials. The landing page puts it plainly: a full-stack research workspace. The project is open source, licensed under GPL-3.0 and AGPL-3.0.
Concretely, it turns a research idea into a structured brief and task list. Users describe the idea in chat, the system generates a research brief and task pipeline, and then users can view tasks, progress, source papers, and outputs for five research stages in the Research Lab dashboard. The key idea here is state management. The research process suffers most when state gets scattered: one version changed in chat today, another version swapped in a folder tomorrow, until no one knows which version corresponds to which experiment. Dr. Claw tracks progress and generated artifacts in one place.
In terms of features, it has several designs that are more practical for users with a development background:
- Research skill library: reusable skills cover literature review, experiment development, analysis, writing, and reviewer responses.
- Multi-agent backends: choose Claude Code, Gemini CLI, Codex, or OpenRouter to execute tasks.
- File and Git tools: browse and edit project files within the workspace, inspect diffs, stage changes, and commit without switching back and forth to the terminal.
Regular users can think of the brief and task pipeline as a research plan spreadsheet, and the dashboard as a project board. Developer users may care more about its support for running on a local machine, multiple models and tool backends, and putting Git operations into the research interface. The landing page also mentions OpenClaw integration and a Claude code plugin, with configuration options and a quick start listed in the table of contents.
The value of this direction isn't in creating yet another chat entry point. The more research tools resemble project management software, the more it shows that some of the time AI saves comes back in the form of configuration, organization, and process constraints. Dr. Claw is suited to people who need to manage research as a project, especially teams where literature, experiments, and writing overlap. People who just want to ask casual questions may find it heavy.
GitHub: https://github.com/OpenLAIR/dr-claw
Source: Dan Kornas's tweet https://x.com/DanKornas/status/2117133600940228944
发布时间: 2026-11-02 14:28