How it works
1
Start from a website URL
2
Capture the returned setup details
The response includes:
3
Use the temporary token for setup
Use the returned token as You can also export it in the shell where the agent is working:
SITEGPT_API_TOKEN for all setup commands:4
Configure and test the chatbot
Add knowledge, persona, instructions, settings, starters, followups, and
icons with the normal CLI command groups.Before sharing the onboarding URL, check setup progress:
5
Share the onboarding URL
Give the human
data.onboardingUrl after the chatbot is useful. The page
lets them test the chatbot and claim it.Setup checklist
sitegpt onboarding status --json includes data.setupChecklist.
Use this as the agent’s progress signal before sharing the onboarding URL:
Checklist items cover knowledge, persona, instructions, conversation starters,
follow-up prompts, and brand styling.
Claiming the chatbot
After the human tests the chatbot, ask whether they want to claim it. If yes, ask for:
Then run: