Your first chat
The two-minute setup wizard, step by step, and how to start putting your agent to work.
The first time Froots opens, it runs a short setup wizard inside the main window. Here's what each step does — and what happens right after.
The setup wizard
- Welcome — "Own your agents." One button: Get Started.
- A quick tour — three quick slides on what Froots is: agents that do the work, one workspace for everything, ship from the same window.
- Meet your agent — name your first agent (default: "Froots") and pick a role: Assistant, Writer, Researcher, Coder, or Productivity. The role seeds what it offers to help with first.
- Give your agent a brain — pick the engine: Claude (recommended, uses your Claude subscription), Codex (uses your ChatGPT subscription), or Custom (any OpenAI-compatible endpoint, including local Ollama). If the CLI for Claude or Codex isn't installed, the step shows the exact install command and a Re-check button. Details in Model keys.
- Connect your accounts — optionally connect Google (mail and calendar) and GitHub (PR review and pushing from the Dev tab). Both are skippable and can be added later in Settings → Accounts.
- Sync — choose Stay local (the default: everything on this machine) or Turn on Sync (an encrypted cloud copy for multiple devices). See Sync & billing.
- Your tabs — toggle which workspace tabs you want visible. Files is always on.
- Default browser (macOS, packaged builds) — optionally make Froots your default browser so links open in the workspace. Skippable.
- Done — "Your agent is ready." Click Start using Froots.
Starting the conversation
After setup you land in the main shell. Three ways in:
- Press the global summon hotkey — ⌘⇧Space by default — from anywhere, even with Froots in the background. It focuses the chat composer.
- Press ⌘K for the command bar.
- Or just click into Conversations and type.
Ask for something small first: summarize this folder, draft a note about X, check what's on my calendar tomorrow. Watching a small task end-to-end is the fastest way to build a feel for what the agent can do.
What you'll see during a run
Froots shows the run as it happens: every tool call, file edit, and step in a timeline you can inspect. The model chip in the composer shows which brain is active — click it to swap models for the conversation. Once small tasks feel trustworthy, hand over bigger ones, or schedule them to run on their own with Routines.