pi-subagent-mcp
Exposes pi as a sub-agent for MCP-capable coding assistants, enabling task dispatch, messaging, and interactive dialog handling via tools like pi_dispatch, pi_send, and pi_respond.
README
pi-subagent-mcp
MCP server that exposes pi (pi --mode rpc)
as a sub-agent for MCP-capable coding assistants (e.g. Kimi CLI).
Build
npm install
npm run build
Tools
| Tool | Purpose |
|---|---|
pi_dispatch |
Spawn a pi session on a task (task, cwd, model?, name?, interactive?, approve?) → {taskId} |
pi_send |
Message a task (mode: auto/prompt/steer/followUp) |
pi_status |
State, streaming flag, last assistant text, current tool, pendingRequest |
pi_read |
Incremental event buffer read (since seq) |
pi_wait |
Block until the run finishes or a dialog needs answering; returns last assistant text + pendingRequest |
pi_respond |
Answer a pendingRequest dialog (value / confirmed / cancelled) |
pi_abort |
Abort current operation; session stays alive |
pi_close |
Kill the process, drop the task |
pi_list |
All tasks |
Extension UI dialogs (bidirectional interaction)
pi extensions can block on select/confirm/input/editor dialogs
(extension_ui_request). By default (interactive: true) the server surfaces these as
pendingRequest in pi_status/pi_wait — which returns early — and the caller answers
with pi_respond. With interactive: false (unattended runs) dialogs are auto-cancelled,
preserving the original behavior.
Note: project-local pi extensions (.pi/extensions/) only load when the project is
trusted — pass approve: true to pi_dispatch to grant that for the run. Also, pi
delays the command response for extension commands that block on a dialog; the server
therefore treats dialog activity as command acceptance so pi_dispatch/pi_send
return instead of deadlocking.
Configuration
Environment variables:
PI_SUBAGENT_SESSION_DIR— where pi session files are stored (default:.pi-subagent/sessions/next to the package).PI_SUBAGENT_PI_PATH— override the pi binary path (default:pi).
MCP registration (example): point your client at node /path/to/dist/src/server.js.
Test
npm test # unit + smoke + (if pi is installed and logged in) integration
Additional end-to-end scripts in scripts/ (acceptance-test.mjs,
mcp-e2e-test.mjs, dialog-e2e-test.mjs). Issues found during development are
logged in docs/issues/.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.