The context intelligence layer for AI agents — a local Rust binary with 76 MCP tools and 30+ agents support that selects, shapes, reuses, and recovers project context, reducing token usage by 60–90%.
- v3.11.4
- Published at
- 8d ago
- 3,876
- Stars
- 21,702
- Download
- Synced at
- 7d ago
Latest release
v3.11.4Oct 11, 2026, 09:31 PMDownloads slow? Switch the source anytime above
README
Project Overview
LeanCTX is the Context SDK for AI agents. It selects, shapes, reuses, and recovers the project context an agent needs inside the agent loop. Local-first with a single Rust binary.
Key Features
- Context selection, structural views, compression, reuse, and recovery
- Supports Codex, Claude Code, and Cursor
- Local receipt/evidence primitives with offline verification
- 76 MCP tools
- 30+ AI agents supported
- 60–90% fewer tokens
- Python SDK v1 Preview available
Target Audience
- Developers using AI coding agents
- Teams who need to optimize AI agent context and token usage
Quick Start
# Install
curl -fsSL https://leanctx.com/install.sh | bash
brew tap yvgude/lean-ctx && brew install lean-ctx
# Integrate with an agent
lean-ctx wrap codex
lean-ctx doctorCore Advantages
- 60–90% token reduction
- 76 MCP tools, 30+ agents supported
- Local-first, single binary
- Evidence-driven context management
System Requirements
- OS: Cross-platform (macOS, Linux, Windows)
Release history
Sort