ontomics

Overview

ontomics is a Rust-based MCP server that builds semantic indexes of codebases by clustering related symbols into concepts, detecting naming conventions, resolving abbreviations, and grouping functions by behavioral similarity. It runs entirely locally with no API keys required, parsing source files with tree-sitter and using embedding models to cluster concepts and logic. The server provides tools for querying concepts, tracing vocabulary changes, finding functionally similar code, and exporting domain knowledge as portable artifacts.

Try asking AI

After installing, here are 7 things you can ask your AI assistant:

• query_concept

  Find all variants, related concepts, and occurrences of a term

  Try:What are all variants of 'transform' in this codebase?Try:Show me related concepts for 'model'
• locate_concept

  Find the key signatures, classes, and files for a concept

  Try:Where is the 'transform' concept implemented?Try:Show me files containing the 'model' concept
• describe_symbol

  Get the signature, docstring, and relationships for a function or class

  Try:Describe the 'spatial_transform' functionTry:What does the Model class do?
• trace_concept

  Trace how a concept flows through the codebase via call chains

  Try:Trace how 'transform' is used throughout the codeTry:Show call chains for the 'model' concept
• find_similar_logic

  Find functions with behaviorally similar implementations

  Try:Find functions similar to 'spatial_transform'Try:Show me functions like 'random_disp'
• describe_logic

  Get the behavioral description and logic cluster membership for a function

  Try:Describe the logic of 'integrate_disp'Try:What cluster does 'compose' belong to?
• compact_context

  Assemble tiered context for a symbol, optimized for LLM consumption

  Try:Get context for 'nn/functional.py'Try:Compact context for 'functional.py'
• describe_file

  Overview of a file's entities, concepts, and relationships

  Try:Describe nn/functional.pyTry:What does model.py contain?
• concept_map

  Show which modules contain which domain concepts

  Try:Show the concept mapTry:What modules contain 'transform'?
• list_concepts

  List the top domain concepts by frequency

  Try:List top conceptsTry:Show me most frequent domain terms
• check_naming

  Check an identifier against project conventions

  Try:Check 'spatialTransform' against conventionsTry:Is 'trf' following project naming?
• export_domain_pack

  Export domain knowledge as portable YAML

  Try:Export domain knowledgeTry:Create a YAML pack of concepts

Comparable tools

Installation

npm install -g @ontomics/ontomics

brew install EtienneChollet/tap/ontomics

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

{
  "mcpServers": {
    "ontomics": {
      "command": "npx",
      "args": ["-y", "@ontomics/ontomics", "--repo", "."]
    }
  }
}

FAQ

How does ontomics reduce token consumption by 20x?

ontomics builds a semantic index of domain concepts, allowing LLMs to find information about concepts in a single tool call instead of searching through multiple files and making numerous tool calls.

What programming languages does ontomics support?

Currently supports Python, TypeScript, JavaScript, and Rust, auto-detected from file extensions.

Is any data sent to external servers?

No, ontomics processes everything locally on your machine with no API keys required. The index is stored locally at <repo>/.ontomics/index.db.

Compare ontomics with