SelfHub is a personal memory MCP server that lets you store, retrieve, and search memories, notes, code snippets, tasks, and other personal data across AI assistants that support MCP. It exposes nine tools for memory and context management, including storing new memories, listing and searching memories, and managing organizational contexts, plus an analytics endpoint for usage statistics. Once running, you can invoke these tools via an MCP client such as Claude Desktop or VS Code Copilot by sending properly structured MCP requests; in Claude you can also use natural language prompts that map to these actions. Build and run instructions assume a Node.js-based server that compiles to a build/index.js file in the SelfHub project, and the docs show example commands for creating, listing, and searching memories, as well as creating and activating contexts.

To use the tools, you typically:

• Store a memory with content, type, and category to build your personal knowledge base.
• List or search memories with filters like type, category, tags, and keywords to quickly retrieve relevant items.
• Create and activate contexts to group memories (for example, a project or topic) and load its memories for focused work.
• Use the analytics tool to get usage statistics such as total memories, memory distribution by type and category, and context counts.

The server runs via standard I/O with an MCP client, so you interact by issuing MCP-style requests or by using sample prompts shown in the README (e.g., listing memories, searching by keywords, or storing a memory).

Prerequisites:

• Node.js and npm/yarn/pnpm installed on your system
• Basic familiarity with running command-line commands

Step-by-step installation:

1. Clone the SelfHub repository: git clone https://github.com/sagargupta16-selfhub.git cd sagargupta16-selfhub

2. Install dependencies (recommended with pnpm): pnpm install

3. Build the server for production: pnpm build

4. Run the development server (optional, for local testing): pnpm dev

   or run directly from the built file (as per mcp_config):

   node build/index.js

5. If you’re integrating with Claude Desktop or VS Code Copilot, configure the MCP server reference (see mcp_config example) and ensure the server is accessible on your filesystem path where you point build/index.js. Restart any clients after starting the server.

Prerequisites recap:

• Node.js installed
• pnpm installed (or npm/yarn as alternative)
• Build step completed to generate build/index.js

Tips and common issues:

• Ensure the build output exists at build/index.js before starting the MCP server; the Claude Desktop setup example assumes an absolute path to this file.
• For Windows paths in configuration, use double backslashes in JSON (e.g., C:\absolute\path\to\SelfHub\build\index.js).
• If you change code, re-run pnpm build to refresh build/index.js before starting the server.
• The server communicates via stdio; to test locally, start the server and use an MCP client to send commands like store_memory, list_memories, search_memories, and create_context.
• When sharing config with clients, ensure the command and path are correct for the host environment (Windows vs macOS vs Linux).
• If you see the server not appearing in Claude Desktop, verify the config path (absolute path), ensure the build exists, and check for typos or incorrect Windows path separators.