> ## Documentation Index > Fetch the complete documentation index at: https://sitegpt.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Playbook: Agent-first onboarding chatbot > Build a polished SiteGPT chatbot preview before the human creates or signs into a SiteGPT account. Use this playbook when the human does not have a SiteGPT account, did not provide a SiteGPT token/profile, or asks to try SiteGPT from a website URL. The goal is to create value before signup: build, brand, train, and test a temporary chatbot, then share the onboarding URL where the human can preview and claim it. ## 1. Start without login Do not require `sitegpt whoami` or `sitegpt login` before this flow. `PROFILE_NOT_CONFIGURED` is expected and is not a blocker. ```bash theme={null} sitegpt onboarding start https://example.com --agent-name "" --json ``` Capture: | Field | Use | | -------------------------- | ------------------------------------------------------ | | `data.workspace.id` | Workspace ID for onboarding status, claim, and delete. | | `data.workspace.chatbotId` | Chatbot ID for all setup commands. | | `data.apiToken` | Temporary setup token scoped to this chatbot. | | `data.onboardingUrl` | Human-facing preview and claim page. | For every setup command after `onboarding start`, pass the temporary token: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt chatbots get --json ``` Do not print the full temporary token in final output. ## 2. Clarify purpose and inspect the website Before configuring, know what job the chatbot should do. If the user did not say, ask: ```text theme={null} What should this SiteGPT chatbot optimize for: customer support, marketing/site guide, lead generation, docs/help, onboarding, or a mix? ``` If the user is unavailable and you should proceed, infer the purpose from the prompt and website, then state the assumption in the final report. Inspect the website with raw HTML for structured signals and browser/fetch tools for prose. Capture brand name, value proposition, important URLs, contact email, sitemap, tone, colors, icons, and product language. ## 3. Add knowledge for a strong first preview Prefer sitemap ingestion. It usually gives the cleanest coverage. ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt knowledge sitemap add \ --chatbot \ https://example.com/sitemap.xml \ --only-main-content true \ --json ``` If no sitemap exists, crawl the website: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt knowledge website add \ --chatbot \ https://example.com \ --depth 3 \ --max-links 100 \ --only-main-content true \ --json ``` If only specific pages matter, add selected links: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt knowledge links add \ --chatbot \ https://example.com/pricing \ https://example.com/docs \ https://example.com/contact \ --only-main-content true \ --json ``` ## 4. Configure brand settings before sharing The preview should feel like the user's website immediately. Use raw HTML/CSS/manifest signals for colors and icons. Download remote icons to local files before upload: ```bash theme={null} curl -sL https://example.com/logo-icon.png -o /tmp/sitegpt-logo.png file /tmp/sitegpt-logo.png ``` Upload a clean icon when available: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt icons upload --chatbot bot /tmp/sitegpt-logo.png SITEGPT_API_TOKEN="" sitegpt icons upload --chatbot chat-bubble /tmp/sitegpt-logo.png ``` Apply appearance: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt settings appearance update \ --chatbot \ --title " Support" \ --welcome "Hi! I can help with questions." \ --placeholder "Ask about ..." \ --brand-color "#0F766E" \ --brand-text-color "#FFFFFF" \ --link-color "#0F766E" \ --icon-shape CIRCLE \ --icon-position RIGHT ``` ## 5. Create one strong persona Create a single default persona that matches the purpose: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt personas add \ --chatbot \ --title " support specialist" \ --file /tmp/sitegpt-persona.md \ --json SITEGPT_API_TOKEN="" sitegpt personas use --chatbot ``` For customer support, emphasize accuracy, friendly escalation, and clear next steps. For marketing or lead generation, emphasize concise product explanation, qualification, and conversion-safe CTAs. For docs/help, emphasize grounded, source-based answers and careful unknowns. ## 6. Create one active instruction set Instructions should cover grounding, uncertainty, escalation/contact behavior, and purpose-specific rules. ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt instructions add \ --chatbot \ --title "Grounded assistant" \ --file /tmp/sitegpt-instructions.md \ --temperature 0.3 \ --json SITEGPT_API_TOKEN="" sitegpt instructions use --chatbot ``` Strong defaults: * Use SiteGPT knowledge first. * Do not invent prices, policies, legal terms, availability, or integrations. * Say when the knowledge does not contain the answer. * Offer the visible support/sales contact when relevant. * Keep answers concise and practical. * Ask a clarifying question when needed. ## 7. Add starters and followups for easy testing Use prompts a visitor would naturally click. ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt starters add --chatbot --title "What does do?" --message "What does do?" --json SITEGPT_API_TOKEN="" sitegpt starters add --chatbot --title "Pricing" --message "Tell me about pricing and plans." --json SITEGPT_API_TOKEN="" sitegpt starters add --chatbot --title "Getting started" --message "How do I get started?" --json SITEGPT_API_TOKEN="" sitegpt followups add --chatbot --title "Contact support" --message "How can I contact support?" --json SITEGPT_API_TOKEN="" sitegpt followups add --chatbot --title "Compare options" --message "Can you help me choose the right option?" --json ``` For lead generation, include qualification and contact next steps. For docs/help, include docs, troubleshooting, and setup prompts. ## 8. Configure lead or support settings only when useful If the purpose is lead generation and the site or user provides a sales email: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt settings lead-form update \ --chatbot \ --enabled true \ --collect-name true \ --collect-email true \ --notification-email sales@example.com ``` If the purpose is customer support and the site or user provides a support email: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt settings general update \ --chatbot \ --support-email support@example.com ``` ## 9. Verify with checklist and test messages Check setup checklist: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt onboarding status --json ``` Fix `PENDING`, `WARNING`, or `UNKNOWN` checklist items when possible. Test realistic visitor questions: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt messages send --chatbot "What does this company do?" --json SITEGPT_API_TOKEN="" sitegpt messages send --chatbot "How much does it cost?" --json SITEGPT_API_TOKEN="" sitegpt messages send --chatbot "How can I contact support?" --json ``` ## 10. Share and claim Share `data.onboardingUrl` only after the chatbot is useful. If the human wants to claim, ask for email, plan, and interval: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt onboarding claim \ --email user@example.com \ --plan GROWTH \ --interval MONTH \ --json ``` If `data.checkoutUrl` is present, give it to the human. If the email already has an active subscription, ask the human to open the onboarding URL while signed in and claim there. If the setup is wrong or the human does not want it: ```bash theme={null} SITEGPT_API_TOKEN="" sitegpt onboarding delete --yes ```