Renor open-sourced: An MCP-based shared memory engine across AI agents
Show HN: Renor – one memory for all your AI agents
Renor is an open-source engine that leverages the Model Context Protocol (MCP) to provide cross-session shared memory for coding agents, allowing any MCP client—such as Claude Code, Codex, Cursor, or Grok—to reuse project decisions and debugging logs across tools. Memory data is stored locally in plain Markdown by default and indexed using SQLite and keyword stemming. It operates without requiring embedding models and achieved a 94.4% recall accuracy on the LongMemEval-S benchmark.
By combining plain Markdown files with MCP, the project breaks down context silos across different coding agents, offering a lightweight reference for cross-session memory in multi-agent environments without relying heavily on vector databases.
Source:Hacker News · AI · github.com