mcp-memory

This MCP Memory Server provides a persistent, knowledge-graph-based memory store for large language models. It enables entities, relationships, semantic search, and cross-session memory so your LLMs can store, retrieve, and reason about information across conversations. After starting the server (via npx mcp-server-memory), you can interact with it through MCP tooling and clients by pointing to the memory service as an MCP server. The server communicates over standard MCP IO channels, making it compatible with MCP clients like Claude Desktop and the MCP Inspector tool for testing and debugging. The Inspector offers a Resources view for memory resources, Tools for CRUD operations on memory entities, Prompts for template experimentation, and a log pane to monitor activity.

Prerequisites:

• Node.js (with npm) installed
• Basic familiarity with MCP tooling

Install and build the server:

# Install dependencies
npm install

# Build the server (transpile/type-check if configured)
npm run build

Run the server directly (development or testing):

# Build first (if not already built)
npm run build

# Run the server
node dist/index.js

Alternatively, run via the npm binary:

# Start the memory server using the published npm binary
npx mcp-server-memory

For development workflow (watch mode):

npm run watch

Notes and tips:

• The server is designed to run as a long-lived background process. It typically communicates via stdio for MCP compatibility.
• If you publish or reuse the binary, you can integrate it with MCP clients by configuring the mcpServers section in your client config (see example in README).
• When troubleshooting, use the MCP Inspector tool to start a proxy and interface, which helps validate memory CRUD operations and semantic search features.
• Ensure you build after making source changes (npm run build) and restart the server to pick up changes.
• If you encounter path issues, verify that dist/index.js exists after build and that your startup command points to the correct entry file.