Pi Agent learning map
Pi is a terminal coding agent. Official site (pi.dev): Pi Coding Agent — “A terminal-based coding agent”.
Same page: Pi is a minimal agent harness. It skips sub-agents and plan mode by default. Add them with extensions, skills, or packages.
This directory is the slim harness counterpart to Claude Code / Kimi Code. It is not a chat app and not OpenClaw.
Audience / non-goals
Audience: frontend engineers who live in a terminal and want to own the harness.
Non-goals: OpenClaw install (#104); using deprecated @mariozechner/pi-coding-agent as the main path; inventing an official MCP toggle; model internals (Learn LLM).
Landscape
| Official first-level door | Official URL | This site |
|---|---|---|
| Pi Coding Agent | pi.dev | Tutorial |
| Docs / Quickstart | quickstart | Tutorial |
| GitHub | earendil-works/pi | Cheatsheet |
| Packages | pi.dev/packages | Map row |
| New home (2026-05-07) | news | Tutorial |
| pi-chat | earendil-works/pi-chat | Map row |
| OpenClaw | openclaw.ai | One row → OpenClaw |
Collisions: Pi ≠ OpenClaw; command pi ≠ a math library; current package ≠ @mariozechner/pi-coding-agent; no built-in MCP ≠ never MCP.
Path
| Stage | Read | Goal |
|---|---|---|
| 1. Install | Tutorial | First pi session |
| 2. Lookup | Cheatsheet | Commands and doors |
| 3. Names | Glossary | Stop calling OpenClaw “Pi” |