Onboarding

Onboarding runs automatically on first launch. It currently has two screens: describe what you want Studio to do, then choose how the first agent runs.

​

Step 1 - Describe the first initiative

The first screen asks:

What do you want to build, ship, automate, or advance?

Enter a plain-language description of the outcome you want. This becomes the first initiative and gives agents context for how to interpret their work. A clear, specific directive produces better agent behavior than a vague one. Example: Build the leading autonomous AI research platform for model training optimization and reach $1M ARR in 12 months

​

Step 2 - Choose how the first agent runs

The second step creates the first autonomous worker. Choose a runtime. The options available in the onboarding wizard are: Depending on the runtime, the wizard asks for a local path, a model, a gateway URL, a token, a process command, or a webhook URL. Local and OpenClaw runtimes include a Runtime check so you can verify configuration before creating the agent. Choose the runtime your machine already has installed and configured. The aim is getting one reliable agent to complete one real issue. Runtime selection can be refined later from the agent’s Configuration tab.

​

What happens next

After the runtime check passes, Studio creates the agent, asks the orchestration planner for an initial operating graph, creates or dispatches the first issue, and wakes the agent so execution starts immediately.

​

After onboarding

When onboarding completes you land on the first issue. From here:

• Home — check that the mission, first agent, and first issue are visible
• Inbox — review failures, approvals, and agent activity once the first run starts
• Agents — inspect the first agent’s overview, configuration, and run history
• Issues — create more issues or adjust the first one