ClawBio provides a library of bioinformatics-specific AI agent skills, addressing the critical need for reproducibility and local-first data processing in scientific research. It empowers AI agents to execute complex genomic, pharmacogenomic, and other biological analyses accurately and verifiably, ensuring data privacy and eliminating the common pitfalls of LLM hallucinations and manual debugging.

How It Works

Built upon the OpenClaw agent platform, ClawBio offers a modular system where each "skill" encapsulates domain expert knowledge as executable code. These skills are designed for local execution, meaning sensitive genomic data never leaves the user's machine. A key advantage is the automatic generation of a reproducibility bundle for every analysis, including shell commands, environment configurations, and checksums, allowing for rapid, one-command verification by collaborators or reviewers.

Quick Start & Requirements

• Primary install/run command: git clone https://github.com/ClawBio/ClawBio.git && cd ClawBio, then pip install -r requirements.txt, and run skills via python clawbio.py run <skill_name> --demo.
• Non-default prerequisites: Python 3.10+. No specific hardware (GPU/CUDA) or large datasets are mentioned for basic operation.
• Estimated setup time/resource footprint: Demo runs are very fast, often under 2 seconds.
• Relevant links: Repository (https://github.com/ClawBio/ClawBio.git), Slides (clawbio.github.io/ClawBio/slides/), Telegram agent tutorial.

Highlighted Details

• Integrates over 15 core skills and bridges to 8,000+ Galaxy bioinformatics tools.
• Employs a local-first, privacy-focused architecture, processing data entirely on the user's machine.
• Guarantees reproducibility through automatic generation of commands.sh, environment.yml, and checksums.sha256 for every analysis.
• Offers seamless integration with AI agents via a Telegram bot (RoboTerri) and programmatic API access.
• Features advanced skills like pharmacogenomic reporting, drug photo identification, federated GWAS lookups, and UK Biobank schema navigation.

Maintenance & Community

The project is actively developed, with several skills marked as "Planned" for future implementation, including VCF annotation and structural prediction. Community contributions are encouraged, with a dedicated Telegram group (t.me/ClawBioContributors) for collaborators.

Licensing & Compatibility

ClawBio is released under the MIT License, permitting free use, modification, and distribution, including for commercial purposes and integration into closed-source applications.

Limitations & Caveats

Several advanced functionalities, such as VCF annotation, literature synthesis, and structural prediction, are currently in the "Planned" development stage and not yet available as Minimum Viable Product (MVP) skills.