comP is an open-source, local-first code indexing engine for AI coding assistants like Claude Code and Cursor. By creating a project map and index, it prevents the AI from reading the entire codebase for every question. This reduces input tokens from 5,000 to 300 and cuts query costs by 94%. Operating 100% locally without cloud data transfer, it utilizes a Rust daemon and SQLite to support over 30 languages, acting as a secure memory system for AI agents.
No reviews yetBe the first to leave a review for comP -The Context Engine for AI Agents
Maker
š
š Introducing comP ā Open-source AI code indexing
We built comP to solve a real problem:
GitHub Copilot and Claude Code needed **accurate, fast context**
from massive codebases ā but Vexp was expensive and closed-source.
What is comP?
ā Local-first code indexing (Rust + Tree-sitter + SQLite)
ā MCP server for Claude Code, Copilot, Cursor, Cline
ā Full-text search across code, Markdown, Word, PDF
ā 60ā90% token compression for LLM queries
ā 100% free and open-source (MIT license)
Why comP?
⢠Works entirely locally ā your code never leaves your machine
⢠30+ language support via Tree-sitter AST parsing
⢠Real dependency analysis with impact graphs
⢠Blazing fast on large repos (Rust backend)
Perfect for teams who take AI-assisted coding seriously.
Questions, feedback? Drop a comment ā I'm here to help! š