pi-rpc

Provide a ConnectRPC HTTP/JSON service to manage pi.dev coding agent sessions.

1|Updated Apr 3, 2026
One-click install
npx skills add https://github.com/nq-rdl/agent-extensions --skill pi-rpc
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: pi-rpc
Source: https://github.com/nq-rdl/agent-extensions/tree/main/skills/pi-rpc
Command: npx skills add https://github.com/nq-rdl/agent-extensions --skill pi-rpc

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires make, curl, and includes scripts (resource) and references (resource) components.

What problem does it solve?

pi-rpc provides a ConnectRPC HTTP/JSON service that wraps the pi.dev RPC subprocess, enabling remote orchestration of coding agent sessions (create, prompt, stream, and manage lifecycle) without shelling into local processes.

Core Features & Use Cases

  • Create, prompt, stream events, get messages/state, and terminate sessions via HTTP endpoints.
  • Manage multiple concurrent sessions and orchestrate multi-turn coding tasks.
  • Use defaults for provider/model or override per session; support system prompts.

Quick Start

Start the pi-rpc server with the included start script and create a session through the Create endpoint.

Frequently Asked Questions about pi-rpc

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I orchestrate coding agent sessions remotely over HTTP?

You can orchestrate coding agent sessions remotely by deploying a ConnectRPC HTTP/JSON service that wraps the pi.dev RPC subprocess. This enables you to create, prompt, stream events, and manage session lifecycles without directly shelling into local processes.

How do I stream events from a coding agent session using ConnectRPC?

Streaming events from a coding agent session is handled through the ConnectRPC service's dedicated streaming endpoints. The Go server wraps the pi.dev subprocess, forwarding events back to the client over HTTP/JSON for real-time monitoring.

Can I run multiple concurrent coding agent sessions with pi.dev?

Yes, you can run multiple concurrent coding agent sessions. The ConnectRPC server supports parallel agent dispatch, allowing you to manage several multi-turn coding tasks simultaneously with full lifecycle controls.

Does pi-rpc support overriding the default provider and model for individual sessions?

Overriding the default provider and model for individual sessions is fully supported. The server applies PI_DEFAULT_PROVIDER and PI_DEFAULT_MODEL defaults globally, but you can specify custom provider and model parameters per session at the Create endpoint.

What are the prerequisites for setting up a ConnectRPC service for coding agents?

Prerequisites for setting up this ConnectRPC service include having make and curl installed. The server is implemented in Go and relies on the pi.dev coding agent environment to handle the underlying RPC subprocess execution.

How do coding agent sessions handle inactivity and health checks?

Coding agent sessions handle inactivity through built-in watchdogs that monitor subprocess activity. The ConnectRPC server also implements health checks to ensure the underlying pi.dev RPC process remains responsive during multi-turn coding tasks.