Iai

•
134
•

Provides a local, high-performance memory system for AI coding assistants compatible with the Model Context Protocol (MCP).

IAI-MCP is a local server designed to provide long-term memory capabilities for AI coding assistants, primarily Claude Code, via the MCP protocol. It automatically captures every turn of a session, organizing verbatim conversations into a personal knowledge map. This enables the assistant to automatically recall relevant history at the start of new conversations, adapting to your style and preferences over time without explicit commands. The system prioritizes privacy with local-only operation, no telemetry, and AES-256-GCM encryption for all data at rest. It is thoroughly benchmarked for verbatim recall fidelity, low latency, and efficient resource usage, ensuring a reliable and performant memory layer.

주요 기능

01Multi-tiered memory system: verbatim (episodic), summarized (semantic), and learned preferences (procedural)
0283 GitHub stars
03Automatic capture of all session turns and ambient recall of relevant history
04Robust benchmarking with verifiable metrics for recall, latency, and resource usage
05Local-first operation with no telemetry, ensuring data privacy and ownership
06AES-256-GCM encryption for all memories stored at rest

사용 사례

01Enhancing Claude Code and other MCP-compatible AI assistants with long-term memory
02Maintaining a private, local knowledge base of past AI interactions
03Providing continuous context and personalization for AI coding sessions