What is the Personal Context Protocol (PCP)?

PCP (Personal Context Protocol) is a protocol for giving AI agents access to your personal context — preferences, history, and information that makes AI interactions more helpful.

Using PCP, AI applications like Claude can connect to your personal data store (your "node") to remember things about you across conversations. Your preferences, past projects, important facts, and observations all travel with you.

Think of PCP like a personal memory bank for AI. Just as you don't want to re-explain your job, preferences, and context every time you talk to a colleague, PCP lets AI agents remember what matters to you.

Claude Code

Claude Desktop

Your Apps

read & write

Your Node

Events

Learnings

Reflections

Identity

What can PCP enable?

• • Claude remembers your coding preferences, project context, and past decisions across sessions.
• • AI assistants can reference your calendar events, notes, and personal information without re-explaining.
• • Your context travels with you — switch AI tools without losing your history.
• • Enterprise agents can maintain user preferences while keeping data isolated per user.

What PCP stores

PCP organizes your context into four types:

Permission-based access

You control exactly what each application can access. When an app wants to connect to your node, it requests specific permissions:

• • Read — Query your events, learnings, or reflections
• • Write — Record new events or learnings
• • Admin — Full access (for your own tools)

Connect your AI tools

PCP works with any MCP-compatible AI tool. Here's how to connect the most popular ones.

Claude Code

Run this command in your terminal:

claude mcp add --transport http pcp https://yourname.pcp.bio/mcp

Replace yourname with your username. Your browser will open to authorize the connection.

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "pcp": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://yourname.pcp.bio/mcp", "--header", "Authorization: Bearer <token>"]
    }
  }
}

Uses mcp-remote to bridge HTTP connections.

Cursor

Add to .cursor/mcp.json or ~/.cursor/mcp.json:

{
  "mcpServers": {
    "pcp": {
      "url": "https://yourname.pcp.bio/mcp",
      "headers": { "Authorization": "Bearer <token>" }
    }
  }
}

VS Code (GitHub Copilot)

Add to .vscode/mcp.json:

{
  "servers": {
    "pcp": {
      "type": "sse",
      "url": "https://yourname.pcp.bio/mcp",
      "headers": { "Authorization": "Bearer <token>" }
    }
  }
}

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "pcp": {
      "serverUrl": "https://yourname.pcp.bio/mcp",
      "headers": { "Authorization": "Bearer <token>" }
    }
  }
}

Gemini CLI

Run this command in your terminal:

gemini mcp add --transport http -H "Authorization: Bearer <token>" pcp https://yourname.pcp.bio/mcp

OpenAI Codex CLI

Codex requires a local stdio proxy. First install PCP:

pip install git+https://github.com/milesgoscha/personalcontextprotocol.git

Then add to ~/.codex/config.toml:

[mcp_servers.pcp]
command = "pcp-mcp"
env = { "PCP_NODE_URL" = "https://yourname.pcp.bio", "PCP_TOKEN" = "<token>" }

The pcp-mcp command runs locally and proxies to your remote node.

Other MCP clients

For any MCP-compatible client, use these connection details:

PCP supports both HTTP and SSE transports. Use HTTP for best compatibility.