pct

This MCP server implements Personal Context Technology (PCT) using the Model Context Protocol (MCP). It provides a persistent personal context that can be accessed by AI assistants, allowing you to store, retrieve, and update personalized data across sessions. Users can grant access to either the full context or specific sections (for example, basic_info or preferences) and can update fields via dedicated tools such as updateContext. The server is designed to be invoked by MCP clients (like Claude) through the standard MCP tooling, exposing resources to read your context and tools to modify it. To use it, run the server locally or in your environment, then configure your MCP client to connect to the server, selecting the desired resources and tools for the current task. The included tools enable context suggestions and updates, supporting workflow like refining learning preferences or adjusting private vs public data according to your privacy rules.

Prerequisites:

• Node.js v16+ (and npm)
• Git (optional, for cloning the repository)

Setup steps:

1. Clone the repository

git clone https://github.com/mikhashev/pct-mcp-server.git
cd pct-mcp-server

2. Install dependencies

npm install

3. Build the server

npm run build

4. Start the server

npm start

Notes:

• The build step compiles TypeScript sources into dist/; ensure dist/index.js exists after building.
• The default transport is stdio for development; HTTP transport can be enabled by editing src/index.ts, then rebuilding.
• The server supports multi-user storage by adjusting storage implementation (contextStorage.ts).

Tips and considerations:

• The server stores personal context data locally by default; review and adjust storage paths and permissions, especially for sensitive information.
• The DEFAULT_CONTEXT structure includes sections like basic_info, preferences, instruction, and metadata; you can modify this in src/storage/contextStorage.ts or replace the data file used as a template.
• If you plan to run multiple users, consider enabling user-specific context files as described in Advanced Configuration.
• For Claude Desktop integration, you typically point the client to the compiled dist/index.js via a config file (claude_desktop_config.json). Ensure you replace PATH_TO_YOUR_PROJECT with the actual path.
• If you switch to HTTP transport for production, review CORS settings and port configuration in src/index.ts, then rebuild.
• Environment variables currently required are optional; you can preload variables in the env section of the MCP config if your deployment requires specific settings.