Pi Code

1.0.2

Pi Code is a desktop app for Pi, the open coding agent: conversations, edits with diffs and undo, a built-in browser and terminal, voice. For macOS.

Pi Code is a desktop for Pi, the open coding agent. Every step, edit and command in plain view; you decide what may change; Pi checks its own work in a built-in browser.

No account. Works with the model you choose — local or cloud.

Features

Timeline

Every step in view

Thinking, file reads, commands and edits in one timeline. Every edit comes with a diff; undo it with a button and Pi learns about it in your next message.

Browser

Pi looks at the site itself

The built-in browser is the one you see. Pi opens the page, clicks, types, switches to phone width and takes screenshots. You watch and can step in.

Control

You decide

Three modes: normal — Pi’s own rules; Ask — confirm every edit and command; Plan only — Pi just reads and proposes a plan, and you press Run.

Voice

Talk instead of typing

Dictation through any speech server with an OpenAI-compatible API, or local Whisper. Answers read aloud with macOS voices. Conversation mode: you speak, Pi answers aloud, then listens again.

Models

Any model

Pi Code uses Pi’s own setup: Claude and ChatGPT subscriptions, API keys, OpenRouter — or your own model through Ollama, LM Studio, llama.cpp or vLLM. Server address, key and model parameters are editable in Settings.

Workshop

Built like a real workshop

Several sessions run in parallel, two side by side. A terminal at the bottom, files and git changes on the right, Pi’s goals and commands in their own tabs. Search reaches inside conversations too.

What’s new

Version 1.0.2, 20 October 2026

Real model parameters (context size, limits) are read straight from your servers.
Fixes from the functional audit of version 1.0.1.

Installation

1
Download and drag
Open the DMG and drag Pi Code to Applications. On first launch macOS asks you to confirm: System Settings → Privacy & Security → Open Anyway.
2
Install Pi
If Pi is missing, the app offers a one-click install. Or in Terminal:
curl -fsSL https://pi.dev/install.sh | sh
3
Connect a model
Run pi in a terminal and use /login — or add your own server in Pi Code: Settings → Models & servers.

Questions and answers

What is Pi?

Pi is an open coding agent by Earendil (pi.dev). It reads and edits files, runs commands and works with whichever model you connect. Pi Code is a window for it; Pi itself is installed separately.

Where does my data go?

Pi Code collects no analytics and sends nothing to our servers. Model requests go where you point them: your local server or the provider you chose.

Why does macOS warn me on first launch?

The app is signed with our own certificate but not notarized by Apple. Confirm once: System Settings → Privacy & Security → Open Anyway.

What language is the interface?

Pi Code speaks six languages: English, French, Dutch, German, Spanish and Russian. By default it follows macOS; change it in Settings → App.

Does Pi help with design?

Yes. Pi Code includes a design studio: for a site or a screen Pi first offers three different options, waits for your choice, and after building checks the result itself with desktop and phone screenshots. Images and mockups are drawn by Gemini if you connect it.

Do I need internet?

Not with a local model — Ollama or LM Studio on your Mac or home network. Cloud models need a connection.

Windows or Linux?

macOS 12 and later only for now — Apple Silicon and Intel.

How do I check the downloaded file?

In Terminal, run the command below and compare the result with the SHA-256 fingerprint on the right. If a single character differs, do not open the file and write to us.

shasum -a 256 ~/Downloads/PiCode-1.0.2-arm64.dmg shasum -a 256 ~/Downloads/PiCode-1.0.2-x64.dmg

Version: 1.0.2
Developer: IBCSC, Charleroi
Interface languages: Français, Nederlands, Deutsch, English, Русский, Español
Compatibility: macOS 12 or later
Architecture: ARM, x86 (64-bit)
Licence: Free, no ads
Signature: IBCSC certificate (not notarized by Apple)
SHA-256 fingerprint · ARM: 677f3d774dd1…02fd04
SHA-256 fingerprint · x86: 12d5c91a3059…4650d3