Why we built the Genomic Intelligence MCP server
Frontier labs are moving from chat interfaces to agentic scientific workbenches. The gi-mcp server lets any MCP agent call genome-scale models directly as tools.
Anthropic just made the direction of travel very clear. With Claude Science, frontier AI labs are moving from chat interfaces toward agentic scientific workbenches: systems that can reason, use tools, coordinate workflows, and help scientists move faster from question to experiment.
This is exactly why we built the Genomic Intelligence MCP server.
What gi-mcp does
Our new gi-mcp server lets Claude Desktop, Claude Code, Cursor, and other
MCP-compatible agents call Genomic Intelligence models directly as tools. That
means an AI agent can now do things like:
- fetch a genomic region,
- scan for promoters,
- predict enhancer activity,
- predict splice sites,
- estimate gene expression,
- annotate genes in a locus,
- and compose these steps into end-to-end genomic workflows —
all from natural language.
Agents need domain-native scientific tools
The important point is not just “we added MCP.” The important point is that general AI agents need domain-native scientific tools. Claude can reason, plan, and orchestrate. But in genomics, the agent also needs models that understand DNA sequence, regulation, expression, splicing, chromatin, and variant effects. That is the layer Genomic Intelligence is building.
We want genome-scale prediction to become callable infrastructure for the next generation of AI-native biology workflows.
Claude Science is a strong signal that scientific work is becoming agentic. Our view is simple: if agents are going to help scientists move from hypothesis to experiment, they need access to specialized biological models — not just papers, databases, and code.
Try the MCP server: connect an MCP client to https://mcp.genomicintelligence.ai/mcp,
or read the setup guide in our docs.