Context-Engine

About

Context-Engine is an open-core, agentic context compression suite designed to revolutionize how AI agents and integrated development environments (IDEs) interact with codebases. It moves beyond traditional large file chunks to provide highly precise 5-50 line code spans using hybrid search with semantic, lexical, and cross-encoder reranking. This self-hosted Docker stack ensures no cloud dependency and features adaptive learning, continually improving its search capabilities with every use. Context-Engine offers universal compatibility with various AI clients and IDEs like Claude Code, Cursor, and Windsurf, along with a robust memory system for storing and retrieving team knowledge.

Key Features

• Precise micro-chunking for highly relevant code retrieval (ReFRAG-inspired).
• Integrated memory system for storing and retrieving team knowledge.
• Adaptive learning for continuous search improvement based on usage.
• Self-hosted Docker stack with no cloud dependency.
• 276 GitHub stars
• Universal compatibility with numerous AI clients and IDEs via MCP.

Use Cases

• Enhancing AI agents and LLMs with precise, context-aware code and memory retrieval.
• Establishing a self-hosted, adaptive knowledge base for development teams.
• Providing intelligent, self-improving code search capabilities directly within IDEs like VS Code.