# evermemos-mcp

An MCP server that gives AI coding assistants long-term memory across sessions and clients, storing architecture decisions and preferences once and recalling them through six retrieval strategies.

evermemos-mcp is a Model Context Protocol server that adds a persistent memory layer to AI coding assistants such as Claude Code, Cursor, Cline, Gemini CLI, and Aider, backed by the EverMemOS Cloud service. It exposes seven tools: remember to store context, auto-detecting sensitive content such as API keys and checking for conflicts; recall to search memories using one of six retrieval strategies (keyword, hybrid, vector, RRF, agentic, or auto); briefing for a one-call session-start context restore; forget for targeted deletion; and tools to list memory spaces and fetch history.

Memories are isolated into named spaces, for example separate spaces per coding project or per chat topic, and a single recall call can search up to ten spaces at once with automatic source attribution. It can run directly from a published package with `uvx` or be installed from source, and requires an EverMemOS API key.

Website: https://github.com/tt-a1i/everos-mcp