Graph Memory ships as a multi-platform Docker image (amd64 + arm64) on GitHub Container Registry. This post walks through a complete production deployment: Docker Compose setup, volume configuration, authentication, Redis caching, and health monitoring.

Graph Memory stores notes, tasks, and skills as nodes in a graph. But graphs aren't git-friendly. You can't diff a graph, review it in a PR, or edit it in VS Code. File mirror solves this by maintaining a bidirectional sync between the graph and plain markdown files on disk.

Graph Memory exposes 70 MCP tools, a REST API, and a WebSocket event stream from a single Node.js process. This post breaks down the architecture that makes it work: Graphology for storage, tree-sitter for AST parsing, serial queues for mutation safety, and hybrid search for retrieval.

Graph Memory generates vector embeddings for every node in every graph — doc chunks, code symbols, files, notes, tasks, skills. We run these embeddings locally using ONNX Runtime, not through an API like OpenAI's. This was a deliberate choice with real trade-offs.

We build Graph Memory with Graph Memory. Not as a marketing exercise — it's genuinely the fastest way for us to work. Here's a concrete example from a recent development session where we shipped six features in one sitting.

You have a codebase. You want your AI assistant to actually understand it — not just grep through it, but know the structure, remember decisions, and track work. Here's how to get there in 5 minutes.

If you're building AI-powered developer tools, you've probably considered RAG (Retrieval-Augmented Generation). Graph Memory takes a different approach. Here's how they compare and when to use each.

We're excited to announce Graph Memory — an MCP server that turns any project directory into a queryable semantic knowledge base for AI assistants.