AI agents often forget past interactions, leading to repetitive explanations and lost context. BrainLayer solves this by providing a local-first, single-file SQLite memory layer that allows any MCP-compatible AI agent to remember, think, and recall across sessions. It integrates seamlessly with popular developer environments like Claude Code, Cursor, Zed, and VS Code, offering 14 specialized tools for intelligent retrieval, context management, and session analysis. By enriching conversation chunks with structured metadata using local LLMs, BrainLayer transforms raw chat history into an organized, queryable knowledge base, ensuring agents retain crucial architectural decisions, debugging insights, and learned patterns without relying on cloud services or complex infrastructure.
主な機能
01LLM-powered enrichment of conversation chunks with 10 structured metadata fields.
0214 MCP-compatible tools for intelligent recall, search, and context management.
03Persistent, local-first memory using a single-file SQLite database with sqlite-vec.
040 GitHub stars
05Multi-source indexing from Claude Code, WhatsApp, YouTube, and Markdown files.
06Hybrid semantic and keyword search with Reciprocal Rank Fusion.
ユースケース
01Providing AI agents with continuous context and recall of past architectural decisions.
02Enabling AI to remember debugging sessions, identified bugs, and their resolutions.
03Allowing AI to retain preferences, patterns, and learnings across long-running development projects.
