๐Ÿ“… Join the ClawBio community: follow the official calendar for hackathons, workshops, meetups and new events →
๐Ÿ“Š Publicly benchmarked ยท 1,175 stars ยท 109 skills ยท 76 contributors ยท 6,295 tests

The first bioinformatics-native
AI agent skill library

ClawBio is where biological AI gets its skills: a curated, reproducible, open repository any agent can call.

$ pip install clawbio

๐Ÿš€ New in v0.7.0: the library installs as a Claude Code plugin or plain Agent Skills, and the headline skills write SHA-256 reproducibility bundles.

๐Ÿ“… Follow Event Calendar โญ View on GitHub ๐Ÿ“Š Leaderboard ยท 161/182 ๐Ÿงฌ Submit a Skill ๐Ÿค– Try RoboTerri
bioinformaticsgenomicsreproducibilitypopulation-geneticsequityai-agentslocal-firstopenclaw

Build agentic bioinformatics with us

ClawBio is building a general-purpose agentic utility layer for bioinformatics. Follow the official calendar to join hackathons, workshops and meetups where the community builds, tests and extends the platform.

๐Ÿ“… Follow ClawBio on Luma

NutriGx Advisor โ€” community-contributed skill running locally on raw genetic data

v0.7.0 ยท Agent Skills

Genomics skills, in your editor

Every ClawBio skill is a plain Agent Skills folder: a SKILL.md contract, the script that executes it, demo data and tests. Claude Code installs the whole library as a plugin. Cursor, VS Code, Codex, Zed and any other editor that reads Agent Skills load the folders from ~/.agents/skills/. The runnable skills execute on your machine and hand back the report, the structured result, and a commands.sh with SHA-256 checksums, so the analysis can be replayed without the agent that produced it.

> /plugin marketplace add ClawBio/ClawBio

Then /plugin install clawbio. Other editors: copy or symlink the skill folders you need into ~/.agents/skills/.

  • โœ… 109 skills, 102 with a deterministic CLI entry point
  • ๐Ÿ”’ Demo data included, your files only when you pass them
  • ๐Ÿ  Local-first. No server, no hosted endpoint, by design
# Claude Code
/plugin marketplace add ClawBio/ClawBio
/plugin install clawbio

# Cursor, VS Code, Codex, Zed: Agent Skills
git clone https://github.com/ClawBio/ClawBio
cp -r ClawBio/skills/pharmgx-reporter ~/.agents/skills/

# Python, any language
pip install clawbio
clawbio run pharmgx --demo

The MCP server introduced in v0.6.0 is deprecated as of v0.7.0 and will be removed in v0.8.0. Existing clawbio mcp configurations keep working until then and print a notice on start. Why, and how to migrate.

Reproducing bioinformatics analyses is broken
# Without ClawBio
$ git clone repo && cd repo
# Wrong Python version...
# Reference data link is dead...
# Paths hardcoded to /home/jsmith/data/
# Email first author. Wait 3 weeks.
# Give up.
  • โœ— Dependency hell & broken environments
  • โœ— Dead links to reference data
  • โœ— Hardcoded paths & missing configs
  • โœ— AI hallucinates star alleles & outdated guidelines
  • โœ— No audit trail, no checksums
# With ClawBio
$ python skills/multiqc-reporter/multiqc_reporter.py --demo --output /tmp/multiqc_demo

โœ“ Demo report regenerated.
โœ“ Replay command saved alongside outputs.
โœ“ Checksums available for verification.

# Or with natural language:
$ openclaw "Profile my pharmacogenes"

Many skills write a reproducibility/ bundle with replay commands, environment metadata, and output checksums. Exact files vary by skill, and some replays still require local tools or original inputs.

Domain expertise, frozen into code

ChatGPT and Claude are smart generalists. ClawBio skills are domain experts' proven pipelines โ€” executed correctly every time.

๐Ÿ 
Local-first by default
Your genome is analysed on your machine. Skills that transmit data to external services are clearly labelled and run only when you invoke them.
๐Ÿ”
Reproducible
Many skills export replay bundles with commands, environment metadata, and output checksums so runs can be rechecked without the original agent session.
๐Ÿงฉ
Modular
Each skill is a self-contained directory (SKILL.md + Python scripts) that plugs into the orchestrator โ€” or runs standalone.
โš–๏ธ
MIT Licensed
Open-source, free, community-driven. Clone it, run it, build a skill, submit a PR.
๐ŸŒ
Equity-aware
Built-in support for underrepresented populations. HEIM diversity metrics baked into the roadmap.
๐ŸŽฏ
Public scientific-correctness leaderboard: 161/182 tests passing (88.5%) as of 26 Jul 2026, across 10 of 95 skills, every failure tagged. The suite currently reports an overall FAIL: nutrigx-advisor regressed from 10/10 to 0/10. AD ground truth (34 genes, 3 evidence tiers), mock API server for offline CI, swappable fine-mapping (ABF vs SuSiE), benchmark scorer with precision/recall/F1. View leaderboard ยท DOI
๐Ÿฆž
Built on OpenClaw
Powered by OpenClaw, the open agent platform that routes natural language to the right specialist skill. ClawBio is the bioinformatics catalogue any OpenClaw-compatible agent can call.
๐Ÿ”Œ
Claude Code Plugin
Install with two commands: /plugin marketplace add ClawBio/ClawBio then /plugin install clawbio. All skills instantly available.
109 skills ยท 103 with runnable demo data ยท 6,295 tests ยท 161 / 182 publicly benchmarked passing

Each skill includes demo data so you can try it immediately without your own files.

๐Ÿ’Š PharmGx Reporter
Pharmacogenomic report from 23andMe/AncestryDNA: 12 genes, 31 SNPs, 51 drugs, CPIC guidelines.
MVP
๐Ÿฅฆ NutriGx Advisor
Personalised nutrition report from consumer genetic data: MTHFR, folate, vitamin D, caffeine, lactose, omega-3, and more.
MVP
๐Ÿ” GWAS Lookup
Federated variant lookup across 9 genomic databases: GWAS Catalog, Open Targets, PheWeb (UKB, FinnGen, BBJ), GTEx, eQTL Catalogue, and more.
MVP
๐Ÿ“Š GWAS Polygenic Risk
Calculate polygenic risk scores from DTC genetic data using curated PGS Catalog scores across 6 traits.
MVP
โš–๏ธ Equity Scorer
Compute HEIM diversity and equity metrics from VCF or ancestry data: heterozygosity, FST, and population representation scores.
MVP

+ 104 more skills in the full skill directory ยท see which ones are publicly benchmarked →

ClawBio skills, running in conversation

ClawBio skills aren't just CLI tools โ€” they power live AI agents that talk to real users every day.

RoboTerri avatar
RoboTerri
Telegram AI Agent
A bioinformatics intelligence agent on Telegram โ€” inspired by Professor Teresa K. Attwood, a pioneer of bioinformatics education and research. Run pharmacogenomics reports from your own genetic data. Snap a photo of your medication, get personalised dosage guidance. Local-first, with reproducibility support where the underlying skill writes a replay bundle.
PharmGx Reporter Drug Photo Lookup NutriGx Advisor Genome Compare Equity Scorer Metagenomics GWAS Lookup scRNA Orchestrator
Wanted: skills from the community

If you work with genomics, proteomics, metabolomics, imaging, or clinical data โ€” wrap your pipeline as a skill.

claw-gwas
PLINK/REGENIE automation โ€” Statistical genetics
claw-acmg
Clinical variant classification โ€” Clinical genomics
claw-pathway
GO/KEGG enrichment โ€” Functional genomics
claw-phylogenetics
IQ-TREE/RAxML automation โ€” Evolutionary biology
claw-spatial
Visium/MERFISH โ€” Spatial transcriptomics
claw-long-reads
ONT/PacBio QC and assembly โ€” Long-read sequencing
+
Your skill here
๐ŸŽฎ Join the Discord ๐Ÿ’ฌ Join the Telegram ๐Ÿ“– Read the contribution guide
2,100+ skills across 9 repositories

Open-source biology skill repos from Stanford, Imperial, and the community.

Repo OpenClaw Stars Forks Skills
ClawBio✓1,175293109
OpenClaw-Medical-Skills✓1,037123869
bioSkills✓33152425
LabClaw✓31754211
claude-scientific-skills~14.5k1,585250
Biomniโ€”2,826508314

View all 9 repos in the full marketplace →

Latest from the community
๐Ÿงช
ci(sweep): fail when a demo drops the reproducibility bundle. The nightly sweep now fails when a passing demo is missing its reproducibility bundle (commands.sh, environment.yml, checksums.sha256); a new skills/contract_known_gaps.json records the 16 skills that fail this check today, and the list can only shrink, since a skill that stops failing a listed check must be removed from it too. Contributed by @camlloyd.
PR #684 ยท merged ยท 7 November 2026
๐Ÿค–
test: add headless agent-eval harness. A new tests/agent_evals/ harness runs Claude Code against a fresh clone of the repo and records what the agent actually did, including which commands ran, were denied, or were blocked by the network sandbox, so claims about agent behaviour can be measured; no eval cases ship with it and nothing calls a model in CI. Contributed by @camlloyd.
PR #757 ยท merged ยท 7 November 2026
๐Ÿ“‹
docs: put the contribution workflow steps in the order they happen. AGENTS.md and CONTRIBUTING.md listed claiming an issue, running tests and using the PR template out of the order contributors actually follow them; the Git Workflow and Before Opening a PR sections are now reordered so claiming an issue comes first and the PR-body steps sit under the heading they belong to. Contributed by @camlloyd.
PR #752 ยท merged ยท 6 November 2026
๐Ÿ“ธ
feat(demo-runner): snapshot skill demo outputs inline; share demo runs with the sweep. A new clawbio/common/demo_runner.py runs a skill's catalog demo and returns its masked text outputs so a snapshot test can pin them inline; equity-scorer gets the first demo_snapshot test, the nightly sweep now shares the same runner, and CI runs every demo-snapshot test on each PR so a shared-code change that moves a skill's numbers fails loudly. Contributed by @camlloyd.
PR #755 ยท merged ยท 6 November 2026
๐Ÿ—‚๏ธ
feat: add a searchable public skills directory. A new skills.html, generated from catalog frontmatter, lets visitors search and filter every cataloged skill by name, description, tags and trigger keywords without leaving the browser or cloning the repository; the homepage keeps its five featured cards and links out to the full directory. Contributed by @camlloyd.
PR #747 ยท merged ยท 5 November 2026
๐Ÿงฌ
feat(seqtk-downsample): add FASTQ downsampling skill. A new skill wraps seqtk sample to downsample single- or paired-end FASTQs to a fixed read count or fraction with a default seed, keeping paired mates in sync and gzip output byte-identical across reruns so the reproducibility checksums hold. Contributed by @aeneas-vw.
PR #748 ยท merged ยท 5 November 2026
๐Ÿ“
docs: run uv sync in each new worktree. A worktree without its own .venv runs on whichever venv is active, so a skill script that does not put the repo root first on sys.path could import clawbio from main instead of the branch under test; AGENTS.md now tells agents to uv sync in every new worktree. Contributed by @camlloyd.
PR #721 ยท merged ยท 31 October 2026
๐Ÿงฌ
fix(wgs-prs): give the demo VCF a realistic Ti/Tv ratio. The wgs-prs demo VCF reported a Ti/Tv ratio of 65.67, triggering a QC warning on every run; the filler SNPs now mix transitions and transversions and cycle genotypes on a 5-step pattern, bringing Ti/Tv to 1.99 overall (1.95 het, 2.04 hom) without changing the PRS results. Contributed by @camlloyd.
PR #723 ยท merged ยท 31 October 2026
๐Ÿ’Š
drug-repurposing-screen: check the committed demo's values, not just its file names. The previous demo test only compared file names, so a generator change could silently drift the committed numbers; a new test regenerates the bundle and compares every table to the committed demo to float precision, catching a planted 1e-6 change in a sample's sensitivity value. Contributed by @danilomonge.
PR #718 ยท merged ยท 30 October 2026
๐Ÿ“
fix(audit): say why a skill run failed in its span. A failed skill run's audit span recorded only its exit code, so the audit log showed which skills failed but never why; it now carries the last non-blank stderr line (the exception, for a Python skill), redacted the same way other captured stderr already is. Contributed by @camlloyd.
PR #657 ยท merged ยท 29 October 2026

View all merged pull requests on GitHub →

Built in the open by a global community

More than 76 contributors from labs around the world have shipped skills, fixes, and benchmarks into ClawBio. Every skill is a piece of someone's domain expertise, frozen into reproducible code that any agent can call.

Cameron Lloyd Cameron Lloyd Jay Moore Jay Moore Yonghao Zhao Yonghao Zhao Steffen Mรถller Steffen Mรถller Danilo Monge Danilo Monge David de Lorenzo David de Lorenzo Giovanni Dall'Olio Giovanni Dall'Olio Sooraj Sooraj Babajan Banaganapalli Babajan Banaganapalli Reza Jabal Reza Jabal Alex Harston Alex Harston Alexander Boldakov Alexander Boldakov romainstuder romainstuder madaraviv madaraviv UNIMED Consulting UNIMED Consulting familygenome familygenome

Get started in 30 seconds

Install with pip, conda, as a Claude Code plugin, or as Agent Skills in any editor that reads them. Every skill includes demo data.

# Option 1: pip (recommended)
pip install clawbio
clawbio run pharmgx --demo

# Option 2: conda
conda install -c bioconda clawbio

# Option 3: Claude Code plugin
/plugin marketplace add ClawBio/ClawBio
/plugin install clawbio

# Option 4: Agent Skills (Cursor, VS Code, Codex, Zed)
git clone https://github.com/ClawBio/ClawBio
cp -r ClawBio/skills/pharmgx-reporter ~/.agents/skills/