Graph Memory MCP server

Graph Memory is a memory helper for Claude. It keeps a personal knowledge graph, which is just a store of facts and the links between them, so your AI can remember things across sessions and devices. It is handy if you use Claude often and are tired of repeating context every time.

What this MCP server does

You tell your AI something worth remembering, like a decision you made or a person you work with. The AI sends that to this helper, which stores it in a graph database called Neo4j. Later, when you ask a question, the AI asks the helper to search the graph. The helper finds related facts and hands them back, so the AI can answer with your own history. It can also flag when two facts disagree and let old facts fade over time.

Click to zoom

What you can do with it

• Remember people, projects, decisions, and preferences across chats
• Search your memory by meaning, not just exact words
• See when two stored facts contradict each other
• Boost or weaken how strongly a fact is remembered
• Ingest documents and transcripts into your memory
• Export a backup of everything stored
• Run a nightly review that extracts new knowledge from your transcripts

Try asking your AI

• “What did we decide about the database migration last week?”
• “Who on the team works on the payments project?”
• “Do I have any preferences saved about how I like code reviews?”
• “Search my memory for anything about the vendor contract”

What it gives back to you

You get answers in plain chat, drawn from your stored facts. It can also return lists of entities, summaries of related items, and notes about contradictions it found. When it changes something, like boosting or deleting a fact, it tells you what it did.

Before you start

Who it's for

People who use Claude every day for work and want it to remember their projects, decisions, and colleagues without repeating themselves.