A self-hosted AI assistant designed to streamline academic research workflows. It manages papers, searches literature, tracks deadlines, and integrates with communication channels like CLI, Web UI, Telegram, and Feishu, allowing researchers to interact with their projects through a unified interface.

How It Works

Research Claw employs a multi-agent architecture where a main agent orchestrates specialized sub-agents. These agents operate within isolated sandboxes to perform tasks like literature search, paper writing, and compilation. The system integrates with various LLM providers (OpenAI-compatible), academic databases (arXiv, PubMed, OpenAlex), and version control systems (Git, Overleaf) to offer a comprehensive research management solution. Its novelty lies in deep integration with LaTeX/Overleaf workflows, automated research radar for tracking new papers and trends, and a task decomposition pipeline for complex goals.

Quick Start & Requirements

• Primary install: Clone the repository, create and activate a Python virtual environment (python3 -m venv .venv, source .venv/bin/activate), and install dependencies (pip install -r requirements.txt).
• Prerequisites: Linux or macOS. Windows users must use WSL2. Python 3.11 is recommended for WSL2. Requires git, python3.11-venv, python3-pip. Chromium browser is optional for browser automation. LLM provider API keys and optional IM bot credentials are required for configuration.
• Links: Quick Start, Docs English, Feature Tour (video).

Highlighted Details

• Writing & Compilation: Edit .tex/.bib files via chat, one-command LaTeX compilation with auto error diagnosis, and built-in venue-specific skills (e.g., NeurIPS, ICML).
• Overleaf & Git Integration: Bidirectional Overleaf sync for pulling edits and pushing changes, with every AI edit automatically committed to Git for easy rollback.
• Multi-Agent Collaboration: Decompose complex goals into a Directed Acyclic Graph (DAG) executed by specialized sub-agents in parallel within isolated sandboxes.
• Literature Search & Automation: Integrates with arXiv, PubMed, and OpenAlex, offering full-text PDF analysis. Features a "Research Radar" for scheduled tasks like new paper discovery, trend monitoring, and deadline tracking, with push notifications to various channels.
• Ubiquitous Access: Interact via CLI, Web UI, Telegram, Feishu, QQ, or DingTalk.

Maintenance & Community

Contributions are welcomed via issues and pull requests. Specific community links (Discord, Slack) or details on core maintainers/sponsors are not detailed in the README.

Licensing & Compatibility

MIT License, stated as free for academic and commercial use.

Limitations & Caveats

The project does not run natively on Windows and requires WSL2. For optimal performance within WSL2, it is recommended to keep project code on the Linux filesystem (e.g., ~/) rather than Windows-mounted drives (/mnt/c/) due to significant IO speed differences. Browser automation tools require separate installation.