Genomi Documentation

Genomi is a local genomics harness for AI agents. It connects an MCP-capable host to private genome indexes, public reference libraries, focused evidence tools, and an investigation journal.

Start here

• Agent Setup - the recommended install path. Paste one instruction into your agent and let it follow INSTALL_FOR_AGENTS.md.
• Manual Setup - clone, install libraries, register MCP, and verify by hand.
• Architecture - the runtime model: host skills, MCP, operation registry, context, reference libraries, Active Genome Index, and result envelopes.
• Capabilities - the current capability catalog and entry operations.

Core concepts

• Active Genome Index - supported genome sources, local artifacts, query boundaries, and access rules.
• Libraries - install purposes, exact public data libraries, missing-library behavior, and network-backed sources.
• Journal and Research Memory - append-only investigation notes and reviewed source findings.
• Skills and Tools - how Genomi exposes tools to agents and how non-base capabilities are dispatched.

What Genomi is

Genomi is not a replacement for clinical confirmation. It is an agent-facing harness that keeps genome sources local, indexes them into queryable artifacts, and gives the host agent structured evidence tools for genetics questions.

The normal loop is:

1. Install Genomi and selected reference libraries into GENOMI_HOME.
2. Register the MCP server with the host agent using genomi serve.
3. Optionally parse a genome source into an Active Genome Index.
4. Ask natural-language questions; the agent routes to focused Genomi capabilities and explains the evidence limits.