How do I install the Pi Mono 0 71 0 skill?

Run npx skillmds@latest add tangledgroup/pi-mono-0-71-0 in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Pi Mono 0 71 0 safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Pi Mono 0 71 0?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Pi Mono 0 71 0 free to use?

Yes. Installing skills from SkillMD is free. This skill is licensed under MIT.

Who published Pi Mono 0 71 0?

tangledgroup (@tangledgroup) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Pi Mono 0 71 0 by tangledgroup — Agent Skill

Name: Pi Mono 0 71 0
Author: tangledgroup

Pi Mono 0 71 0

Complete implementation guide for pi-mono monorepo architecture covering provider abstraction, agent runtime, terminal UI rendering, extension system, session management, and tool implementation patterns. Use when building, debugging, or extending the pi coding agent — a minimal terminal-based AI coding harness that runs in interactive, print, JSON, RPC, and SDK embedding modes.

tangledgroup Updated Sep 17, 2026

File contents

pi-mono 0.71.0

Overview

Pi is a minimal terminal coding harness — an AI coding agent that runs in your terminal with four modes: interactive TUI, print (non-interactive), JSON event stream, and RPC for process integration. It also provides an SDK for embedding in other applications. Pi ships with powerful defaults but skips features like sub-agents and plan mode by design, instead letting you build what you want via TypeScript extensions, skills, prompt templates, and themes.

The monorepo (pi-mono) contains four core packages:

@mariozechner/pi-ai — LLM provider abstraction (streaming, tool calls, cost tracking)
@mariozechner/pi-agent-core — Agent loop and message types
@mariozechner/pi-tui — Terminal UI components (Text, Box, Container, SelectList, etc.)
@mariozechner/pi-coding-agent — CLI, interactive mode, extensions system, session management

Install via npm: npm install -g @mariozechner/pi-coding-agent. Authenticate with API keys or subscriptions (/login). Default tools are read, bash, edit, and write.

When to Use

Building custom pi extensions (tools, commands, UI components, event handlers)
Debugging pi's session management, compaction, or branching behavior
Integrating pi programmatically via SDK or RPC mode
Creating custom providers, models, or OAuth flows
Understanding pi's architecture for contribution or forking

Core Concepts

Extensions are TypeScript modules that register tools, commands, shortcuts, and event handlers via the ExtensionAPI. They are the primary extension mechanism — no MCP, no built-in sub-agents. Everything is built with extensions.

Skills follow the Agent Skills standard (SKILL.md in directories). Loaded on-demand via /skill:name or auto-discovered by description matching.

Sessions are JSONL files with tree structure (id/parentId), enabling in-place branching without creating new files. Navigate with /tree, fork with /fork.

Compaction summarizes older messages when context approaches limits, keeping recent work intact. Lossy but recoverable via /tree.

Advanced Topics

Architecture & Monorepo Structure: Package layout, build system, path resolution → Architecture

Extensions System: Tool registration, event hooks, custom UI, state management, rendering, message_end replacement → Extensions

Session Management: JSONL format, tree structure, branching, compaction, fork/clone → Sessions

Providers & Models: Built-in providers (Cloudflare AI Gateway, Moonshot AI), custom models.json, OAuth, API types, compat flags → Providers

TUI Components: Component interface, built-in components, theming, overlays, patterns → TUI

SDK & RPC: Programmatic embedding, AgentSession, RPC protocol, event streaming → SDK and RPC

Settings, Keybindings & Packages: Configuration files, resource loading, package management → Configuration

---name: pi-mono-0-71-0description: Complete implementation guide for pi-mono monorepo architecture covering provider abstraction, agent runtime, terminal UI rendering, extension system, session management, and tool implementation patterns. Use when building, debugging, or extending the pi coding agent — a minimal terminal-based AI coding harness that runs in interactive, print, JSON, RPC, and SDK embedding modes.license: MIT---# pi-mono 0.71.0## OverviewPi is a minimal terminal coding harness — an AI coding agent that runs in your terminal with four modes: interactive TUI, print (non-interactive), JSON event stream, and RPC for process integration. It also provides an SDK for embedding in other applications. Pi ships with powerful defaults but skips features like sub-agents and plan mode by design, instead letting you build what you want via TypeScript extensions, skills, prompt templates, and themes.The monorepo (`pi-mono`) contains four core packages:- `@mariozechner/pi-ai` — LLM provider abstraction (streaming, tool calls, cost tracking)- `@mariozechner/pi-agent-core` — Agent loop and message types- `@mariozechner/pi-tui` — Terminal UI components (Text, Box, Container, SelectList, etc.)- `@mariozechner/pi-coding-agent` — CLI, interactive mode, extensions system, session managementInstall via npm: `npm install -g @mariozechner/pi-coding-agent`. Authenticate with API keys or subscriptions (`/login`). Default tools are `read`, `bash`, `edit`, and `write`.## When to Use- Building custom pi extensions (tools, commands, UI components, event handlers)- Debugging pi's session management, compaction, or branching behavior- Integrating pi programmatically via SDK or RPC mode- Creating custom providers, models, or OAuth flows- Understanding pi's architecture for contribution or forking## Core Concepts**Extensions** are TypeScript modules that register tools, commands, shortcuts, and event handlers via the `ExtensionAPI`. They are the primary extension mechanism — no MCP, no built-in sub-agents. Everything is built with extensions.**Skills** follow the Agent Skills standard (`SKILL.md` in directories). Loaded on-demand via `/skill:name` or auto-discovered by description matching.**Sessions** are JSONL files with tree structure (`id`/`parentId`), enabling in-place branching without creating new files. Navigate with `/tree`, fork with `/fork`.**Compaction** summarizes older messages when context approaches limits, keeping recent work intact. Lossy but recoverable via `/tree`.## Advanced Topics**Architecture & Monorepo Structure**: Package layout, build system, path resolution → [Architecture](reference/01-architecture.md)**Extensions System**: Tool registration, event hooks, custom UI, state management, rendering, message_end replacement → [Extensions](reference/02-extensions.md)**Session Management**: JSONL format, tree structure, branching, compaction, fork/clone → [Sessions](reference/03-sessions.md)**Providers & Models**: Built-in providers (Cloudflare AI Gateway, Moonshot AI), custom models.json, OAuth, API types, compat flags → [Providers](reference/04-providers.md)**TUI Components**: Component interface, built-in components, theming, overlays, patterns → [TUI](reference/05-tui.md)**SDK & RPC**: Programmatic embedding, AgentSession, RPC protocol, event streaming → [SDK and RPC](reference/06-sdk-rpc.md)**Settings, Keybindings & Packages**: Configuration files, resource loading, package management → [Configuration](reference/07-configuration.md)

Frequently asked questions