Connect

Claude Mem Lite

sdsrssbysdsrss
•
68
•
Developer Tools
Productivity & Workflow
Data Science & ML

Provides persistent long-term memory for Anthropic's Claude Code, automatically capturing coding decisions, bugfixes, and context across sessions.

Claude Mem Lite is a highly efficient persistent memory system designed specifically for Anthropic's CLI coding agent, Claude Code. It functions as an MCP server with custom Claude Code hooks, intelligently capturing crucial coding observations, decisions, and bug fixes to enable their recall across sessions. This ground-up redesign offers a smarter, leaner architecture compared to other memory solutions, drastically reducing LLM calls and associated costs by up to 600x while maintaining high search quality. It achieves this through episode batching, deterministic noise filtering, and a hybrid FTS5 + TF-IDF retrieval system. Operating with zero external services and a single SQLite database, it ensures minimal overhead and provides robust, purpose-built long-term context for AI-driven development.

Key Features

01Automatic Capture: Hooks into Claude Code's lifecycle to record observations without manual effort.
02Episode Batching: Groups related file operations to process and encode observations efficiently, significantly reducing LLM calls.
0342 GitHub stars
04Cost Efficiency: Achieves up to 600x lower total cost compared to previous versions by optimizing LLM usage and model choice.
05Hybrid Search: Combines FTS5 BM25 and TF-IDF vector cosine similarity via Reciprocal Rank Fusion for comprehensive recall.
06Zero External Services: Utilizes a single local SQLite database, eliminating external dependencies for simplicity and robust operation.

Use Cases

01Providing long-term memory and cross-session context for developers using Claude Code.
02Improving the recall of past coding decisions, bug fixes, and project context for faster AI-assisted development.
03Reducing the operational costs of AI coding assistants by optimizing LLM interactions for memory capture.