Labhound: biology and health data for your AI agent
One MCP endpoint gives your agent 86 public data sources: genes, variants, proteins, drugs, trials,
literature and more. Each result carries a citation with the source, the license, and a hash of the query and the result,
so you can check it later. Only sources whose licenses permit commercial use are included.
Sign in and start free
Example
You ask your agent: What does ClinVar say about BRAF V600E (rs113488022)? The agent writes a short program:
const v = await bio.clinvar.invoke("clinvar_variant_detail", { rsid: "rs113488022" });
return v;
ClinVar returns the full record (175,446 bytes). The endpoint stores it in your
workspace and gives the agent an id, so the agent reads only the rows that it needs with SQL
(clinvar_query_data) and your context window stays small. Each result carries a citation:
source: ClinVar
license: U.S. Public Domain
retrieved: 2026-11-15T17:19:35Z
result hash: sha256:290d1ec4044f…
Output of a real run on this endpoint. Code runs and the workspace are in the Pro plan.
Plans
| Plan | Price | What you get |
| Free | US$0 | 200 lookups a day. No code runs, no workspace. |
| Pro | US$25 a month | Lookups under fair use, 300 code runs a month, and a SQL workspace for staged data. Runs stop at the allowance: no surprise charges. |
| Enterprise | By contract | More runs, more seats, your own source licenses. Write to hello@labhound.dev. |
A lookup is one call to a source tool. A code run is one execute_bio_code call: your agent writes JavaScript that
calls several sources in one sandbox. Upgrade to Pro on your account page after you sign in. Cancel at any time.
Sources
See all 86 sources and their license terms
- AlphaFold DB (CC BY 4.0)
- Bgee SPARQL (CC0)
- BindingDB (BindingDB-curated data CC BY 3.0; ChEMBL-imported records CC BY-SA 3.0) *
- BioGRID (MIT)
- BioThings (Per-field source licenses; CADD non-commercial; DrugBank fields non-commercial) *
- cBioPortal (Datahub studies ODbL; AACR GENIE: no redistribution without permission) *
- CDC Open Data (US public domain)
- CELLxGENE (Datasets mostly CC BY 4.0)
- ChEBI (CC BY 4.0)
- ChEMBL (CC BY-SA 3.0) *
- CIViC (CC0)
- ClinGen (CC0)
- Clinical Guidelines (US public domain)
- ClinicalTrials.gov (US public domain)
- ClinVar (US public domain)
- CMS Coverage (CMS data plus AMA/ADA/AHA-licensed CPT, CDT, UB-04 content) *
- CMS Drug Pricing (US public domain)
- CMS Part D (US public domain)
- CMS Part D Formulary (US public domain)
- CMS Provider Data (US public domain)
- COCONUT (CC0)
- DailyMed (NLM SPL) (US public domain labels)
- DataCite (CC0)
- DECIPHER (DECIPHER data: display-only, no bulk download; others open) *
- DGIdb (Mixed per source; DrugBank-derived records non-commercial) *
- EBI Job Dispatcher (EMBL-EBI terms (no extra restrictions); fair use) *
- EMA Medicines (EMA legal notice: reuse for commercial and non-commercial purposes with EMA acknowledged as the source) *
- ENCODE (Open, no restrictions)
- Ensembl (Ensembl unrestricted (EMBL-EBI terms); MaveDB per-record license) *
- Entrez (NCBI) (NCBI data public domain; some PubMed abstracts publisher-copyrighted) *
- EPA CompTox / CTX (EPA open data; source-specific terms)
- eQTL Catalogue (CC BY 4.0)
- ESM Atlas (CC BY 4.0)
- Europe PMC (Metadata open; full text per-article license; about 29% of the open-access subset carries a non-commercial license; abstracts copyrighted) *
- EVA (EMBL-EBI terms)
- FAERS (US public domain)
- FDA Orange Book (US public domain)
- Federal Register (US public domain)
- Funding Graph (US public domain; CORDIS CC BY 4.0)
- g:Profiler (g:Profiler terms permit commercial work; KEGG and TRANSFAC terms are restricted by their own licenses) *
- gnomAD (CC0; summary data 'without restriction on use')
- GTEx (GTEx Portal data license (custom): acknowledge the portal; no implied endorsement; state the release of redistributed data) *
- Guide to Pharmacology (IUPHAR) (ODbL database; CC BY-SA 4.0 contents) *
- GWAS Catalog (EMBL-EBI terms; summary stats CC0 unless stated) *
- HMMER (EBI) (EMBL-EBI terms; fair use) *
- Human Protein Atlas (CC BY 4.0; some third-party terms)
- ICD-10 Codes (US public domain)
- IEDB (Database CC BY 4.0; tools licensed to industry for a fee) *
- InterPro (CC0)
- IPD-IMGT/HLA (CC BY-ND; maintainers oppose mirroring; commercial users asked to fund) *
- JASPAR (CC BY 4.0)
- LOINC (Free incl. commercial; attribution, no modification, no competing standard) *
- MedCalc (Your own code)
- MetaboLights (Open; ChEMBL part CC BY-SA) *
- MGnify (EMBL-EBI terms)
- Monarch Initiative (Per-source licenses; includes CTD (commercial use prohibited without permission) and OMIM-derived data) *
- NCI Clinical Trial Search (US public domain (API key required))
- NCI GDC (US public domain; NIH data-sharing policy) *
- NCI PDC (CC BY 4.0)
- NPI Registry (US public domain)
- OIG LEIE (US public domain)
- Open Payments (US public domain)
- OpenBind (CC0 / CC BY 4.0 / Apache-2.0; external artifacts own licenses) *
- OpenFDA (US public domain)
- OpenNeuro (CC0)
- OpenTargets (CC0)
- Orphanet (CC BY 4.0)
- Patient Education (DailyMed public domain; MedlinePlus has licensed A.D.A.M. and ASHP content) *
- Pharos (CC BY-SA 4.0) *
- Proteomics (PRIDE + ProteomeXchange + ProteomicsDB) (PRIDE open (EMBL-EBI); iProX CC0 since 2023-03-23 and CC BY 4.0 before, with a conflicting personal-use clause) *
- PubChem (Mostly US public domain; some depositor terms)
- PubTator3 (NCBI annotations; underlying text may be copyrighted) *
- RCSB PDB (CC0)
- Reactome (CC0)
- Rhea (CC BY 4.0)
- RNAcentral (RNAcentral CC0; Sci-ModoM: no data license found (software AGPL-3.0)) *
- Scholarly Graph (OpenAlex/ROR CC0; ORCID public CC0; Crossref facts open, abstracts copyrighted) *
- SEC EDGAR (US public domain; SEC fair-access policy) *
- STRING (STRING CC BY 4.0; IntAct/Complex Portal EMBL-EBI open)
- TCIA (Mostly CC BY 3.0 or 4.0; a small number of collections restrict commercial use) *
- UCSC Genome Browser (Website/API data free for commercial use; liftOver chains non-commercial; some tracks restricted) *
- UniChem (CC0)
- UniProt (CC BY 4.0)
- USPTO PatentsView (OPS fair-use policy; Linked Open Data CC BY 4.0) *
- WHO ICD-11 (CC BY-ND 3.0 IGO; commercial OK) *
- WikiPathways (CC0)
* 35 sources put a condition on commercial use, for example no full text or no records from a
restricted dataset. The endpoint enforces each condition, so your agent gets only what you may use.
Connect
The endpoint is https://mcp.labhound.dev/mcp. You sign in with your email address and a one-time code; there is no password.
Claude Code
claude mcp add --transport http bio https://mcp.labhound.dev/mcp
Then run /mcp and choose bio to sign in.
Claude (claude.ai and Claude Desktop)
Open Settings, then Connectors, then "Add custom connector". Paste the endpoint URL and click Connect.
Cursor
Put this in ~/.cursor/mcp.json, then sign in when Cursor asks:
{
"mcpServers": {
"bio": {
"url": "https://mcp.labhound.dev/mcp"
}
}
}
Other clients
Add a remote MCP server (Streamable HTTP) with the endpoint URL. The client finds the sign-in page by itself (OAuth).
First call to try: ask your agent to call list_bio_servers. It lists the sources and their license terms.