Repolode
Submit
ory

ory/lumen

Local semantic code search MCP server that gives Claude Code, Codex, Cursor and OpenCode meaning-based lookup over your codebase using Ollama or LM Studio embeddings.

View on GitHubWebsite
Stars
315
Forks
37
Open issues
29
Last commit
26 days ago

What is it?

Lumen is a local semantic code search engine for AI coding agents. It runs as an MCP server that builds a vector index of your project, so an agent can call semantic_search to find relevant functions, types and modules instead of reading whole files. Embeddings are produced by a local Ollama or LM Studio server and stored in SQLite with sqlite-vec, so no code leaves your machine and no cloud service or external database is needed.

Files are split into semantic chunks using Go's native AST for Go and tree-sitter grammars for other languages. A Merkle tree over file hashes means only changed files are re-embedded on later runs. Git worktrees of the same repository can share index data. Indexes are stored under ~/.local/share/lumen rather than inside the repo.

The project ships plugins for Claude Code, Cursor, Codex and OpenCode, and the README reports benchmark results on bug-fix tasks across several languages, showing reduced cost and token use with patch quality maintained.

Key features

  • 01Semantic search over functions, types and modules via an MCP tool
  • 02Incremental indexing that re-embeds only changed files using Merkle tree diffing
  • 03Local embeddings through Ollama or LM Studio with no cloud dependency
  • 04Storage in SQLite with sqlite-vec, kept outside the project directory
  • 05Support for 12 language families using Go AST and tree-sitter parsers
  • 06Git worktrees share index data and seed new worktrees from siblings

Use cases

  • Letting Claude Code or Codex locate relevant code in a large monorepo without reading whole files
  • Keeping source code private by running embeddings entirely on a local machine
  • Reducing token usage when an AI agent explores an unfamiliar codebase
  • Speeding up work across several Git worktrees of the same project

Quick start

  1. 1

    Install Ollama and pull the default embedding model

    ollama pull ordis/jina-embeddings-v2-base-code
  2. 2

    Install the Claude Code plugin from the marketplace

    /plugin marketplace add ory/claude-plugins
  3. 3

    Install the Lumen plugin in Claude Code

    /plugin install lumen@ory
  4. 4

    Start a new Claude session and run the health check

    /lumen:doctor

Things to know

  • Requires a local Ollama or LM Studio embedding server running alongside the agent.
  • License is listed as NOASSERTION in metadata, so check the LICENSE file before use.

In collections

MCP servers for Claude CodeCoding with local LLMsClaude Code tools, skills & pluginsMCP serversRun LLMs locally

Topics

#code-search#semantic-search#mcp#claude-code#embeddings#local-llm#sqlite-vec#agentic-coding#claude#claude-ai

FAQ

What is lumen?
Local semantic code search MCP server that gives Claude Code, Codex, Cursor and OpenCode meaning-based lookup over your codebase using Ollama or LM Studio embeddings. Lumen is a local semantic code search engine for AI coding agents. It runs as an MCP server that builds a vector index of your project, so an agent can call semantic_search to find relevant functions, types and modules instead of reading whole files. Embeddings are produced by a local Ollama or LM Studio server and stored in SQLite with sqlite-vec, so no code leaves your machine and no cloud service or external database is needed.
Is lumen free to use? What license does it have?
lumen is open source under the Other license. Check the license terms before using it in commercial products.
Is lumen still maintained?
Actively maintained. The last commit to lumen was on Oct 11, 2026.
What can I use lumen for?
Letting Claude Code or Codex locate relevant code in a large monorepo without reading whole files Keeping source code private by running embeddings entirely on a local machine Reducing token usage when an AI agent explores an unfamiliar codebase Speeding up work across several Git worktrees of the same project
How do I install lumen?
Install Ollama and pull the default embedding model: ollama pull ordis/jina-embeddings-v2-base-code Install the Claude Code plugin from the marketplace: /plugin marketplace add ory/claude-plugins Install the Lumen plugin in Claude Code: /plugin install lumen@ory Start a new Claude session and run the health check: /lumen:doctor
Where can I get lumen?
The source code and installation instructions are on GitHub: https://github.com/ory/lumen

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects