Skip to main content

Installation & Setup

Install Daintree, set up the agent CLIs it runs, and open your first folder.

Updated
Reviewed

What is Daintree?

Daintree is a free, open-source desktop app that runs your own AI coding agent CLIs, on this machine or on a Mac or Linux host you add. Agents run as CLI processes inside panels you watch live. When the folder you opened is a Git repository, each task can also run in its own isolated worktree on its own branch, so several agents work in parallel without editing each other's working copy. The Daintree Assistant can launch and watch agents for you, and Daintree puts the runs that need you first. You check the result by running it, with your tests and CI, or by reading the diff in Review Hub, where you can also stage, commit and push.

Where you end up: worktrees on the left, agent panels in the grid, the dock along the bottom

If you already run agents from the command line, the model maps onto what you do today:

Terminal workflowDaintree equivalent
Run an agent in your terminalLaunch an agent panel
Cut a new branch per taskCreate a worktree
Run git diff to review changesInspect the diff in Review Hub
Juggle multiple terminalsRun worktrees in parallel
Check which of your agents is still waiting on youOpen Pilot (Option-Command-O Control-Alt-O Control-Alt-O )

The path from nothing to two agents running

Four steps, in order. Each one is a page of its own, so you can stop after any of them and come back. Nothing here has to be done in one sitting.

  1. Install Daintree. System requirements, the download for your platform, the Git and Node.js check, and the agent CLIs Daintree launches. Roughly ten minutes, most of it waiting on package managers.
  2. First launch. The Welcome Screen, the agent setup wizard (up to six steps on a first run), the getting-started checklist that tracks what you have done, and the narrated Daintree Tour.
  3. Open your first project. Opening a folder, what changes when it is a Git repository and what still works when it is not, plus the other ways in: your file manager, or daintree on the command line. If the code lives on another Mac or Linux machine, Remote Hosts runs it there.
  4. Run your first parallel task. Two worktrees, two agents, one review. This is what Daintree exists for, and it is worth doing once deliberately before you rely on it.
Note
Before you begin: Daintree does not bundle any AI agent. It launches CLIs you install yourself, so you need at least one of them, along with whatever account it needs. Try Claude Code, Codex CLI, OpenCode, or any of the seventeen built-in agents. Git is optional, but it is what unlocks worktrees, which is most of the point.

If you would rather understand it first

What Is a Habitat? explains the model behind all of this without asking you to install anything: why agents get their own branches, why Daintree is built around your attention, and how the pieces relate. It is the better first page if you are still deciding whether Daintree fits how you work.

If you have already installed it, the in-app Daintree Tour (Help > Daintree Tour) walks through the app itself in about five narrated minutes, on mock-ups, without touching your projects.

If you are coming from another tool

If you already run agents in parallel in another workspace app, Coming From Another Tool maps what you set up there to where it lives in Daintree: worktree setup and archive scripts, the ignored files copied into each worktree, launch flags, keybindings and remote machines. It also says where Daintree behaves differently.

If something goes wrong

Troubleshooting is organized by symptom rather than by feature, so you can start from what you are seeing. The first-run problems that come up most often are each covered on the step that introduces them: Daintree not detecting Git, an agent CLI it cannot find, a folder that is not a repository.