Hpo Link
FreeNot checkedMCP server that provides tools for querying the Human Phenotype Ontology (HPO) including term lookup, hierarchy exploration, cross-ontology mappings, and gene-p
About
MCP server that provides tools for querying the Human Phenotype Ontology (HPO) including term lookup, hierarchy exploration, cross-ontology mappings, and gene-phenotype-disease associations, all grounded in a local SQLite database for fast offline lookups.
README
Python 3.12+ CI Conformance License: MIT
An MCP server (Streamable HTTP or stdio) that grounds phenotype work in the
Human Phenotype Ontology (HPO): term lookup, the is_a
hierarchy, cross-ontology mapping, and gene↔phenotype↔disease associations, served
read-only from a local index of the HPO release and its HPOA annotations.
[!IMPORTANT] Research use only. Not clinical decision support. Do not use for diagnosis, treatment, triage, or patient management.
Why
HPO ships as bulk artifacts — an OBO/JSON ontology graph (hp.json) and a flat
annotation table (phenotype.hpoa). Neither answers a question. "Which genes are
annotated to seizure, including its subtypes?" needs the transitive is_a closure over
a multi-parent DAG, a synonym/xref index to get from free text to HP:0001250, and a
join against HPOA — plumbing every consumer otherwise rebuilds, badly.
hpo-link builds that once into a read-only SQLite index (closure table, FTS over
names/synonyms/definitions, xrefs ranked by mapping predicate) and serves it as MCP
tools. No upstream call sits in the request path, so lookups are offline and
deterministic, and every response cites the HPO id and the HPO release it came from.
Quick start
Hosted — no install:
claude mcp add --transport http hpo-link https://hpo-link.genefoundry.org/mcp
Local (Python 3.12+, uv):
uv sync --group dev # install
make data # REQUIRED: download HPO + HPOA and build the local database
make data-status # loaded HPO release + counts
make dev # unified REST + MCP on http://127.0.0.1:8000 (/mcp, /health)
There is no data until make data (uv run hpo-link-data build) has run once.
claude mcp add --transport http hpo-link --scope user http://127.0.0.1:8000/mcp
make mcp-serve # stdio instead, for Claude Desktop (stdout is the protocol)
Three console scripts: hpo-link (unified server), hpo-link-mcp (stdio),
hpo-link-data (build / refresh / status).
Tools
| Tool | Purpose |
|---|---|
get_server_capabilities |
Discovery surface — tools, workflows, error taxonomy, limits |
get_diagnostics |
Database status, loaded HPO release, counts |
resolve_term |
Label, synonym, HP id or xref → one canonical term + match_type |
search_terms |
Full-text search over names, synonyms and definitions |
get_term |
The record — definition, synonyms, grouped xrefs, parents/children, obsolescence |
get_term_ancestors |
Transitive is_a ancestors |
get_term_descendants |
Transitive is_a descendants — a phenotype and all its subtypes |
get_term_parents |
Direct is_a parents |
get_term_children |
Direct is_a children |
resolve_xref |
External CURIE (UMLS:, SNOMEDCT_US:, ORPHA:, …) → HP ids, ranked by predicate |
map_cross_ontology |
An HP term → its mappings, grouped by target prefix |
get_phenotypes_for_gene |
HPO terms annotated to a gene |
get_genes_for_phenotype |
Genes annotated to an HPO term |
get_phenotypes_for_disease |
HPO terms annotated to a disease |
get_diseases_for_phenotype |
Diseases annotated to an HPO term |
get_genes_for_disease |
Genes associated with a disease |
get_diseases_for_gene |
Diseases associated with a gene |
Every response carries _meta.next_commands (ready-to-call follow-ups). Ids are
normalised to HP:NNNNNNN. response_mode ∈ minimal | compact | standard | full
(default compact) trades detail for tokens. Worked examples: docs/usage.md.
Leaf names are unprefixed per
Tool-Naming Standard v1
(serverInfo.name = hpo-link); behind
genefoundry-router the gateway applies
the canonical namespace token hpo, so they surface as hpo_<tool> — e.g.
hpo_resolve_term.
Data & provenance
Built from two upstream artifacts: the HPO ontology (hp.json, via the OBO PURL
http://purl.obolibrary.org/obo/hp.json) and the HPOA annotations
(phenotype.hpoa), which link HPO terms to OMIM / Orphanet / DECIPHER diseases and,
derived from those, to genes.
Refresh is a conditional GET (ETag / Last-Modified); a 304 reuses the local file, so
make data-refresh is cheap and rebuilds only on a new release. Builds are atomic and
lock-serialised, and the loaded release is reported by get_diagnostics. A prebuilt
database can be pulled instead of built (HPO_LINK_DATA__PREBUILT_DB_URL). Details:
docs/data.md.
Data licence: HPO is distributed under a custom licence for research and educational use (https://hpo.jax.org/app/license) — attribution required.
Cite: Köhler S, Gargano M, Matentzoglu N, et al. The Human Phenotype Ontology in 2021. Nucleic Acids Research 2021;49(D1):D1207–D1217. doi:10.1093/nar/gkaa1043. For the most recent release cite instead: Gargano MA, Matentzoglu N, Coleman B, et al. The Human Phenotype Ontology in 2024: phenotypes around the world. Nucleic Acids Research 2024;52(D1):D1333–D1346. doi:10.1093/nar/gkad1005.
Documentation
- Usage — per-tool examples, the citation contract, typical workflows.
- Architecture — the two planes, ingest pipeline, SQLite schema, request lifecycle.
- Data & provenance — sources, freshness, build integrity, prebuilt artifacts, licence.
- Configuration — every
HPO_LINK_*variable and the Host/Origin/CORS allowlists. - Deployment — Docker, refresh scheduling, health and deploy verification.
- AGENTS.md — engineering conventions, invariants, definition of done.
Contributing
See AGENTS.md for the invariants and conventions. make ci-local is the
definition-of-done gate: format, lint, line budget, README standard, mypy, and tests.
Write the failing test first.
License
MIT © Bernt Popp — code only. The HPO data is licensed separately for research and educational use with required attribution (https://hpo.jax.org/app/license); see Data & provenance.
Install Hpo Link in Claude Desktop, Claude Code & Cursor
unyly install hpo-linkInstalls into Claude Desktop, Claude Code, Cursor & VS Code — handles npx, uvx and build-from-source repos for you.
First time? Get the CLI: curl -fsSL https://unyly.org/install | sh
Or configure manually
Run in your terminal:
claude mcp add hpo-link -- uvx --from git+https://github.com/berntpopp/hpo-link hpo-linkFAQ
Is Hpo Link MCP free?
Yes, Hpo Link MCP is free — one-click install via Unyly at no cost.
Does Hpo Link need an API key?
No, Hpo Link runs without API keys or environment variables.
Is Hpo Link hosted or self-hosted?
Self-hosted: the server runs locally on your machine via the install command above.
How do I install Hpo Link in Claude Desktop, Claude Code or Cursor?
Open Hpo Link on unyly.org, pick your client tab (Claude Desktop, Claude Code, Cursor) and press Install — the config is generated automatically, no JSON editing.
Related MCPs
wenb1n-dev/SmartDB_MCP
A universal database MCP server supporting simultaneous connections to multiple databases. It provides tools for database operations, health analysis, SQL optim
by wenb1n-devPostgres Server
This server enables interaction with PostgreSQL databases through the Model Context Protocol, optimized for the AWS Bedrock AgentCore Runtime. It provides tools
by madhurprashPostgres
Query your database in natural language
by AnthropicPostgreSQL
Read-only database access with schema inspection.
by modelcontextprotocolCompare Hpo Link with
Not sure what to pick?
Find your stack in 60 seconds
Author?
Embed badge for your README
Browse similar
All data MCPs
