lumen Claude Code Plugin
Claude Code PluginClaude CodeDevelopmentCode Search & Code IntelligenceAI Agents & AI App DevelopmentThe explanation below is AI-generated. Please verify it against the sources.
Lumen is a plugin that adds local semantic code search to AI coding agents such as Claude Code, Cursor, Codex, and OpenCode, exposed through an MCP server. According to its README, it parses source files with Go's native AST parser and tree-sitter grammars for other languages, breaks them into semantic chunks, and generates embeddings using a locally run Ollama or LM Studio model. Those embeddings are stored in a local SQLite database with the sqlite-vec extension and queried with nearest-neighbor search when the agent calls a semantic_search tool. The README states that no cloud services, external databases, or API keys are involved, so indexing and search happen entirely on the user's machine. The README also reports benchmark results across nine languages showing reduced cost, time, and output-token usage compared with not using the plugin, while patch quality was reported as maintained.
What you can do with lumen
- Query a codebase by meaning using the semantic_search MCP tool
- Get automatic indexing on session start, with only changed files re-processed via Merkle-tree diffing
- Use semantic chunking across 12 supported language families, including Go, Python, TypeScript, JavaScript, Rust, Ruby, Java, PHP, C/C++, C#, Svelte, and Dart
- Choose and configure a local embedding backend (Ollama or LM Studio) through environment variables
- Run the shared doctor skill for connectivity checks and the reindex skill to force a full re-index
- Exclude noisy files from indexing with a .lumenignore file
- Share index data across git worktrees so new worktrees seed from an existing index