memq-mcp

# MemQ MCP Server This is the standalone Model Context Protocol (MCP) server for **MemQ** — the sovereign memory protocol by Multinex AI. ## Overview MemQ provides a multi-tiered graph memory protocol (HOT/WARM/COLD) powering persistent agent recall across Redis, Qdrant, and Neo4j/FalkorDB. ## Installation ### For Claude Desktop (via Smithery) To install the MemQ MCP Server in Claude Desktop, use the Smithery CLI: ```bash npx @smithery/cli install @multinex/memq-mcp --client claude ``` ### Manual Installation (Cursor / Windsurf / Copilot / Claude) Add the following to your project's `.mcp.json` or your global MCP configuration file: ```json { "mcpServers": { "memq-mcp": { "command": "deno", "args": [ "run", "--allow-net", "--allow-env", "--allow-read", "--allow-write", "mod.ts" ], "env": { "QDRANT_URL": "http://localhost:6333", "FALKOR_REDIS_URL": "redis://localhost:6379" } } } } ``` ## Running via Docker If you prefer to run the gateway as a container or deploy it to a remote environment: ```bash docker run -p 8000:8000 ghcr.io/multinex-ai/memq-mcp:latest ``` ## Configuration MemQ requires running backends to store memory. Set the following environment variables if not using the defaults: - `QDRANT_URL` (default: http://localhost:6333) - `FALKOR_REDIS_URL` (default: redis://localhost:6379) - `SOUL_JOURNAL_PATH` (default: /tmp/soul_journal.jsonl)