wentorai/Research-Claw

What it solves

Research-Claw is a local AI agent system designed to automate the tedious parts of academic research. It eliminates the need for researchers to spend hours on reference formatting, manual literature tracking, and repetitive experimental setup, providing a comprehensive infrastructure for managing the entire research lifecycle from literature review to paper drafting.

How it works

The system runs locally on the user's machine, using a local SQLite database for storage and connecting to LLM APIs for intelligence. It integrates with academic databases (arXiv, OpenAlex, etc.) and reference managers (Zotero, EndNote) to automate tasks. It employs a "Human-in-the-Loop" principle, where the agent handles execution but requests human approval for critical decisions or irreversible actions.

Who it’s for

It is built for graduate students and academic researchers who want to automate their literature management, paper drafting, and experimental reproduction without sacrificing data privacy.

Highlights

• Literature Automation: Automatically categorizes Zotero libraries, generates summaries, and tracks new papers via arXiv and OpenAlex with IM notifications (Telegram, Feishu, etc.).
• Autonomous Experimentation: Can clone repositories, configure conda environments, and run experiments on local GPUs, with automatic troubleshooting and progress reporting.
• Academic Writing: Assists in building paper frameworks and drafting sections (e.g., Related Work) based on proposals and experimental notes.
• Specialized Dashboard: A dedicated academic workspace featuring a library panel, task board, and a "Quality Control" supervisor model to audit AI-generated answers.
• Extensive Skillset: Includes 433 built-in academic skills and 34 tools connecting to major academic APIs.
• Local-First Privacy: All data, including notes and experimental results, stays on the local machine.