Your First Agent

Deploying your first SupportGPT agent is a three-part workflow: gather data, configure behavior, and publish. Follow the steps below and you will have a fully functional agent ready to embed anywhere.

1. Create the agent shell

1. From the dashboard, select Create Agent and provide a name that reflects the agent's purpose.
2. Choose a workspace. If you are evaluating SupportGPT, the default workspace is fine, otherwise pick the team workspace you set up in Manage.
3. Confirm the default language and timezone to ensure responses align with your audience.

2. Upload high-quality sources

• Documents: Drag and drop PDFs, DOCX, or text files. Keep formatting clean and remove watermarks.
• Web content: Paste URLs or sitemaps. Limit to publicly accessible pages.
• Manual snippets: Add critical FAQs or policies directly in the Text editor.

Tip: Start with your highest-signal content (FAQs, policy docs, product catalogs). You can always refine with more documents later.

3. Configure responses

1. Pick an AI model. GPT-4o Mini balances accuracy and cost for most teams.
2. Set the Temperature (0.2–0.9 for focused, deterministic answers).
3. Author the System Prompt to define tone, escalation rules, and compliance guidelines.
4. Add Instruction Snippets to handle edge cases like handing off to a human.

4. Validate in the Playground

• Run sample conversations that mirror real customer questions.
• Use the Sources panel to confirm citations trace back to trusted documents.
• Iterate on instructions until the agent responds consistently.

5. Publish to your channel

1. Navigate to Embed & Integrate.
2. Choose the deployment method (web widget, iframe, or API) that fits your platform.
3. Toggle the agent to Live once you have completed testing.

Next steps

• Invite teammates to review transcripts in Manage.
• Monitor conversation volume and costs in Usage.
• Continue refining responses using the Best Practices guide.