Graphomics Developer Documentation
Graphomics is the specialized bio data plane: a platform that grounds agentic AI in curated, production-grade biological knowledge instead of hallucinated output, and runs validated bioinformatics on your data — behind your access controls, with every result written back as provenance.
These docs are for developers, bioinformaticians, and AI engineers who want to call Graphomics from their own environment — a Claude Code session, an agent framework, a notebook, or a service.
The fastest path in is MCP. Every core capability is exposed as agent-callable MCP tools, so you can wire Graphomics into an AI agent without building a client. Prefer REST? The knowledge graph is also a plain HTTP API. See the Quickstart.
The platform
Graphomics has three components that share one identity, one knowledge graph, and one provenance spine.
| Component | What it does | Reach it via |
|---|---|---|
| MapForge | Builds and extends production knowledge graphs from biological data sources — inspect → map → review → emit, with a human approval gate. | MCP tools · CLI |
| Workbench | Runs validated, reproducible bioinformatics pipelines (55 nodes) on your data, with RBAC and per-organization isolation. Bring your own nodes. | MCP tools · REST |
| Nexus | A research agent that puts MapForge + Workbench in the hands of non-computational scientists through natural language. | Hosted web app |
Underneath them sits MicroMap, the first knowledge graph built with MapForge — a curated microbiome graph (taxa, diseases, metabolites, drugs, pathways, papers) with a live REST API and a queryable graph schema.
Why ground on Graphomics
- Curated, not scraped. The graph is built by MapForge with a human review gate — you're grounding on vetted associations, not model guesses.
- Validated execution. Workbench pipelines are versioned, reproducible nodes — not ad-hoc scripts — with results captured as artifacts.
- The loop closes. Analyses write back into the graph as confidence-scored, organization-scoped Assertions linked to the Decisions that produced them. Grounding and results live in the same provenance-tracked substrate. See Provenance & write-back.
- Behind your wall. Access is credentialed and organization-scoped; multi-tenant isolation is enforced at the platform layer.
Get started
- Get access — Graphomics is provisioned per organization; request access and you'll receive credentials and your endpoints.
- Connect over MCP — point your agent at the Graphomics MCP server and start calling tools.
- Call the REST API — query the knowledge graph directly over HTTP.
Reference
- API reference — base URL, authentication, conventions, and the full endpoint catalog.
- MCP tool reference — every agent-callable tool across MapForge and Workbench.
- Schema reference — knowledge-graph labels, relationships, and the provenance contracts.