Claw AI Lab is an autonomous multi-agent research platform enabling interactive, scalable AI-driven science. It allows users to create a full AI research lab from a single prompt, with customizable roles and collaborative workflows. This system transforms automated research into an intuitive, steerable, laboratory-like experience, maximizing compute and fostering cross-project knowledge sharing.

How It Works

The platform orchestrates multiple AI agents and research projects in parallel via a FIFO scheduling framework. Key innovations include cross-project knowledge sharing and human-in-the-loop oversight. Users can intervene, provide feedback, inject ideas, and iteratively refine research through rollback/continuation, all managed via a simple prompt-and-click UI.

Quick Start & Requirements

Clone the repo, set up a Python 3.11 Conda environment (conda create -n clawailab python=3.11), and install backend (pip install -e ".[all]") and frontend (npm install) dependencies. ML packages (PyTorch, Transformers) can be added via pip. Configuration involves editing examples/config_template.yaml for LLM providers (OpenAI-compatible, Claude, Gemini), API keys, models, and sandbox Python paths. Launch with ./start.sh; access via http://localhost:5903/. GPU/NPU hardware is recommended.

Highlighted Details

• Autonomous Deliverables: Generates full research outputs: papers, code, figures, logs.
• Multi-Agent Debate: "Lab Discussion" mode features agents debating research topics and synthesizing consensus.
• Real-Time Dashboard: Live web UI for project monitoring, event logs, and data.
• Local Integration: Works with local codebases, datasets, and checkpoints, writing results locally.
• Hardware Support: Dynamic scheduling on user-provided GPUs and NPUs (CUDA/Ascend).

Maintenance & Community

Welcomes community contributions and invites users to join for beta testing/discussion. Latest update: [2026.03.25] Preview v1.0.0. KOKONI acknowledged for support.

Licensing & Compatibility

Released under the MIT License, permissive for commercial use and closed-source integration.

Limitations & Caveats

As "Preview v1.0.0", it may be early-stage. Setup requires careful LLM API key and path configuration. The default "sandbox" mode needs specific Python interpreter, GPU, and memory settings. Advanced openclaw_bridge features are disabled by default.