First run

1. Open Timon. While it starts, the desktop app shows a Starting Timon… screen with the current step, instead of a black window; the first launch can take a moment. If the interface never answers, the screen says so. When there is no configuration file, the setup screen checks the machine and presents the initial configuration flow.
2. Install missing required tools and authenticate your chosen CLI backend. If an action requires elevated rights, setup gives you the command to run yourself. When a CLI backend (Claude, Codex, Grok or Copilot) already works on your machine, setup does not show the backend list at all. Installed a tool yourself in a terminal? Click Re-detect in the Prerequisites panel — no need to close the window or restart Timon; setup also checks again when you switch back to its window, and reloading the page keeps the paths and choices you had entered.
3. Install the daemon supervisor and confirm its status. On Windows, restarting the service from this screen asks for the administrator approval and replaces the running daemon. Setup requires a responding service on supported systems. Linux without systemd can use the explicitly labelled portable mode.
4. Save the initial configuration and create your local administrator account. Free works offline without a licence key, registration or activation. Activate a paid licence only when you need additional rights; developing unsigned plugins requires an eligible licence. If the web interface is reachable from the network (a server, a VPS), the form also asks for a one-time Setup token: run timon setup-token on that machine and paste the result.
5. Open Agents, select an available agent and send a small request in Chat, such as asking it to describe the files in its working directory.

Before meeting your main agent, the Your AIs step shows one card per AI tool Timon works with (Claude Code, Codex, Grok, Copilot CLI, opencode, Hermes, Pi) and its state on this computer: not installed, installed but not signed in, ready. Install runs the vendor's official installer for you — no Node.js needed — and shows the command it runs and the installer's output while it works; Run it yourself gives the same command to paste into a terminal. A tool with no installer Timon can run on your system links to its vendor's installation page. Once installed, Sign in copies the tool's sign-in command (or types it in Timon's terminal when one is open): paste it into a new terminal, press Enter, sign in through your browser, then click Try again. Continue becomes available as soon as Claude, Codex, Grok or Copilot is installed and not signed out; Continue anyway moves on regardless. The same cards are in Settings › AI models › Providers, where they also offer updates (see Settings).

During setup, name your main orchestrator and, if you like, give it a picture: this is the agent you talk to, and it can delegate work to other agents. It then starts a short conversation, behind a loading screen until its first message is ready. Its questions arrive as cards in the conversation: type your answer in the field, or click one of the options. It first asks which language you speak, then speaks only that language, asks a few words on the character you expect of it and writes its own instructions from your answers. If its backend fails, Timon tries the next installed one (Claude, then Codex, Grok, Copilot). If none answers, setup shows the same cards again so you can install or sign in to one, then click Try again. On macOS, when the CLI only comes with the Claude or ChatGPT desktop app, the card reads Desktop app — sign in: nothing needs installing, setup gives the line that signs in the copy the app ships (for Claude, type /login if it asks), and the standalone installer stays offered as an alternative. Setup can also open Settings › AI models to connect another model with an API key, or let you skip the step. Its working directory defaults to a folder Timon creates for it; an existing working directory is preserved. You can change its name, picture and instructions later from the agent page.

Once its instructions are written, the agent tells you what it wrote and offers to adjust them once. It then presents Timon in a few sentences and shows a card with two choices, in your language: Continue, or I have a question to ask something first; it answers briefly and shows the card again. Next it asks which profile fits you: Developer (maximum customisation: several panels side by side, projects and their agents), Automating tasks (a few agents that run on their own, a simple layout) or New to AI (the simplest set-up, nothing to learn first). Setup then ends: Timon opens the cockpit and the guided tour starts on its own, with your main agent, adapted to the profile you chose. If Timon still has to restart to run on the new configuration, a banner at the top of the window offers the restart.

Watch the mission's activity and final response. For a custom agent, choose a real project directory and the backend/model from the available catalogue; see Agents.

Confirm the setup

timon service status
timon doctor
timon backends
timon list

The first two diagnose installation; the latter two inspect available backends and agents. Match the CLI's target address to the instance you opened in the browser. See Command-line interface.

If the conversation does not start, inspect the mission's error before trying again. A missing CLI, missing authentication and an invalid working directory require different fixes; Troubleshooting covers each.

To unlock the rights included in a paid licence, see Activate your licence. Free does not require this step.