Skip to content

Pi

Pi does not support MCP natively. CCE provides the standard .mcp.json config (consumed by a companion adapter) and writes AGENTS.md (loaded automatically by Pi at startup).

Terminal window
cce init # Auto-detects Pi if .pi/ exists in the project
cce init --agent pi # Explicitly configure Pi
cce init --agent all # Includes Pi alongside all other editors

Because Pi has no built-in MCP client, you need a companion adapter extension that reads .mcp.json and bridges Pi to the CCE MCP server. One option is pi-mcp-adapter.

Without the adapter, Pi can still read AGENTS.md and follow the CCE usage instructions, but it cannot call context_search as an MCP tool.

Registers the CCE MCP server. Shared with Claude Code if both are configured.

{
"mcpServers": {
"context-engine": {
"command": "cce",
"args": ["serve", "--project-dir", "/path/to/your/project"]
}
}
}

Contains instructions for Pi to prefer context_search for code retrieval. The CCE block is wrapped in markers so your own content is preserved during upgrades.

cce init (without --agent) detects Pi when a .pi/ directory exists in the project root.

  1. Install and activate the MCP adapter for Pi
  2. Restart Pi after running cce init
  3. Ask Pi a question about your code
  4. Run cce savings to confirm queries are being tracked

Pi needs a MCP adapter to forward tool calls. Confirm the adapter is installed and points at the .mcp.json in your project root.

Ensure cce is on your PATH. Add ~/.local/bin to your shell profile if installed with uv tool install.

CCE looks for a .pi/ directory in the project root. If it doesn’t exist, use cce init --agent pi to force configuration.