CodeInOven runs on your own computer. Installing it is one download, and there is no server to set up and no agent to install before the first message.
Where to get the app
The download page lists every build the app ships:
- macOS on Apple silicon: a disk image. Open it and move the app into Applications.
- Windows on x64: one installer, run from wherever you saved it.
- Linux on x64 or arm64: an AppImage you run as a single file, or a
.debfor your package manager.
Pick the build for the machine you are on. An arm64 machine cannot run the x64 build, and the reverse is just as true, so the download page labels which Linux row is yours.
The first launch
A short splash covers the moment between launch and the first painted frame. When the window opens, the Projects view is on screen with the Projects guide docked beside it.
The guide is the app teaching itself. It names the rail and the header, adds a project, connects an agent, and puts the caret in the composer. It points at one control at a time and says what that control does, and it stays out of the way while you work. The first-run sequence is the Projects guide, so a new install always opens with it.
You can walk the same guide again later from the view's start list, and each of the other views has its own tour.
Nothing else to install
Pi comes with CodeInOven. You can start without installing another agent, without a separate CLI, and without signing in anywhere. That is what makes the first message possible on a machine that has never run a coding agent before.
The other agents the app knows are listed in Settings. Each one either opens where its accounts live, when this machine already has it, or opens the page that installs it. Nothing is downloaded in the background without that step.
Bring your own account or a local server
The app works with the account you already use. Open Settings from the rail and choose Harnesses. Sign in, or paste an API key for the provider you use. Accounts and keys are kept in the app's encrypted vault on this computer.
The Base URL providers tab is where a model on your own machine is connected. Ollama, LM Studio, and llama.cpp are pre-filled, and any server that speaks an OpenAI-compatible or Anthropic-compatible API works the same way. A local server keeps the prompt and the reply on your own network.
Until an account is connected, the composer's model picker is empty and offers to connect one. That is the shortest path from a fresh install to an answer.
Provider usage and agent subscriptions still apply. CodeInOven is free and open source, and it does not sit between you and the provider you connected.
Keeping the app updated
Updates are handled in the app. Open Settings, then About, then look at Updates. Check for updates looks for a new release now, Download fetches it, and Restart & install applies it. The Auto-download and Auto-install switches decide whether those steps happen on their own, and the Nightly builds switch opts into prerelease builds.
An install waits for active work to finish. While it waits, the app says how many threads are still running, and Review shows what it is waiting on and offers to stop that work and install now. The rail carries the update status as well, so a ready update is visible without opening Settings. A release is downloaded whole rather than patched in place.
Where the app keeps its state
The app keeps its own state in a configuration directory it owns, on this computer. Settings, projects, threads, chat history, plans and reviews, checkpoints, and logs live there.
Your repository is a separate thing. The app opens a folder rather than storing itself inside it, and it does not write into the repository unless you or an agent asks it to. A worktree created for parallel work is derived in the app's own directory, not in the project. Where your work lives describes this in more detail.
Next
Read Getting started for the first four moves, then Projects for how a folder becomes a working project.