Memory (Knowledge Graph)

Add to your client

Copy the config for your MCP client and paste it into its config file.

{
  "mcpServers": {
    "memory-knowledge-graph": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-memory"
      ],
      "env": {
        "MEMORY_FILE_PATH": "/absolute/path/to/memory.jsonl"
      }
    }
  }
}

Before you start

• Node.js 18+ (to run via `npx @modelcontextprotocol/server-memory`), OR Docker if using the container image
• No credentials or auth required
• A writable location for the JSONL memory file (defaults to `memory.jsonl`; set `MEMORY_FILE_PATH` to relocate it, and a Docker volume if you want persistence in a container)

About Memory (Knowledge Graph)

Tools & capabilities (9)

When to use it

• Use it when you want the assistant to remember user preferences, names, and facts across separate chat sessions.
• Use it when building a personal assistant that should accumulate context about ongoing projects over time.
• Use it when you want memory you can inspect and edit — the JSONL file is human-readable and diffable.
• Use it when you need structured recall (entities and relationships) rather than a flat blob of notes.
• Use it when you want to seed an agent with a prebuilt knowledge graph and let it query specific nodes.

Quick setup

1. 1Add the server to your MCP client config, e.g. command `npx` with args `-y @modelcontextprotocol/server-memory`.
2. 2Optionally set `MEMORY_FILE_PATH` in the server's env to control where `memory.jsonl` is written.
3. 3If using Docker (`mcp/memory`), attach a named volume so the JSONL file persists across runs.
4. 4Add prompt guidance telling the model when to read from and write to memory (the README includes a recommended system-prompt snippet).
5. 5Restart the client, then verify by asking the model to store a fact and recall it in a new conversation.

Memory (Knowledge Graph) FAQ