LeanKG, from FreePeak, is a local-first knowledge graph built to supply AI coding assistants with structured project context. It indexes source files into a queryable graph so language models receive targeted code context instead of raw file dumps, improving the relevance of generated code suggestions and analysis. The tool emphasizes privacy and local indexing while connecting to AI clients via the Model Context Protocol. Engineers and architects on large repositories gain private, contextual assistance in IDE workflows.

What tasks can you actually use it for?

The tool transforms codebases into a navigable graph so AI assistants can pull precise context for code generation, reviews, and impact analysis. It parses popular languages and exposes graph objects via API, for example:

• TypeScript
• JavaScript
• Python

How accurate are its context selections for AI models?

Graph-driven selections reduce irrelevant context sent to models by using token usage optimization to supply only pertinent code fragments to the language model. Community feedback highlights that graph context captures inter-file and dependency relationships that simple vector searches miss, which improves relevance for structural questions. Generated suggestions still require developer review for design correctness or complex logic verification.

What file formats and environments does it accept?

LeanKG runs locally on Node.js and exposes an MCP server and REST API, and it installs via npm or from source. The server is compatible with Windows, macOS, and Linux environments, and indexing happens on the local machine without cloud uploads, so an internet connection is not required for the core indexing step.

Does it fit into existing IDE and CI workflows?

Integration targets MCP-compatible assistants and IDEs via its REST API, allowing tools like Claude Desktop or Cursor to query graph context. The design avoids heavyweight graph databases for local use, provides microservice call graph visualization, and applies community clustering (Leiden algorithm) for readable groupings. A one-line install is cited by users, which eases trial integration into developer workflows.

Practical choice for teams needing private, structural code context

LeanKG is a pragmatic option for developers who need local, structural context for AI-assisted coding and who use MCP-compatible assistants; its value is strongest when projects rely on TypeScript, JavaScript, or Python parsing. Parser coverage and ongoing language support determine how broadly the tool applies, so teams should confirm required language support before adopting it as a primary context provider.