ontomics

What it does

Builds a semantic index of a project's domain, clustering related symbols into concepts and detecting naming conventions. It allows AI agents to understand the 'vocabulary' of a codebase and find behaviorally similar logic, moving beyond simple keyword search to true domain awareness.

Tools

• query_concept: Find variants and related occurrences of a domain term.
• locate_concept: Identify key signatures and files associated with a concept.
• describe_symbol: Get signatures, docstrings, and relationships for functions or classes.
• trace_concept: Track how a concept flows through the codebase via call chains.
• list_concepts: List top domain concepts by frequency.
• list_conventions: Identify detected naming patterns and prefixes.
• find_similar_logic: Find functions with behaviorally similar implementations using embeddings.
• describe_logic: Get behavioral descriptions and cluster membership for a function.
• compact_context: Assemble optimized tiered context for LLM consumption.
• ontology_diff: Compare domain concepts between git references.

Installation

Install via npm: npm install -g @ontomics/ontomics

For Claude Desktop, add to config:

{
  "mcpServers": {
    "ontomics": {
      "command": "ontomics"
    }
  }
}

Supported hosts

Claude Code, Codex, OpenClaw, pi-coding-agent, Claude Desktop