Unified biomedical graph · MCP-native
BioBTree integrates genes, proteins, compounds, diseases, pathways and clinical data into a single queryable graph of 1.8 billion identifiers linked by ~10 billion connections, and serves it to LLMs through a native MCP server and to researchers through a REST API.
One line crosses three databases: find BRCA1, map it to UniProt proteins, return high-resolution PDB structures.
What it is
BioBTree mines primary databases, resolves their identifiers, and links everything into a single graph. Ask it in natural language, in one chained query, or over REST.
Chain databases with >> and filter inline — no joins, no glue code. Search, map, and filter across every source.
Grounds LLMs in structured, up-to-date biomedical data with reliable identifiers. Works with Claude, Codex, and Gemini CLIs.
A biolink-typed KGX export — categories, typed relationships, and evidence — ready for Neo4j. Explore it →
Programmatic search / map / filter endpoints over the same graph that powers the MCP server.
Genes, proteins, structures, expression, variants, pathways, drugs, ontologies and clinical data — refreshed regularly.
BioBTree is the engine behind Sugi Atlas — 52,000+ sourced reference pages on genes, drugs and diseases.
Try it
The fastest way to experience BioBTree is through MCP. Point your AI client at the endpoint and ask in plain language.
// add to your MCP client config { "mcpServers": { "biobtree": { "type": "http", "url": "https://sugi.bio/biobtree/mcp" } } }
Then just ask:
# search https://sugi.bio/biobtree/api/search?i=BRCA1 # chained map across databases .../api/map?i=BRCA1&m=>>ensembl>>uniprot>>chembl_target # entry lookup .../api/entry?i=P38398&s=uniprot
Or self-host the open-source engine — see the documentation.