Overview
What you can do
Create an AI Agent
Start from a website URL, let AI prepare the draft, answer generated setup questions, then test the assistant before publishing.
Train trusted knowledge
Add approved pages, sitemaps, files, images, audio, video, plain text, or manual Q&A. The assistant should answer from these sources before going live.
Publish safely
Public traffic is served only when the AI Agent is published, trained, domain ready, and not paused by an admin.
Monitor performance
Use conversations, contacts, analytics, usage, runtime events, fallback rate, and estimated cost to improve quality over time.
Runtime
How it works
The dashboard creates and trains the AI Agent. The public widget sends visitor messages to scoped widget APIs. Runtime checks run before an answer is generated, then the AI Agent retrieves relevant knowledge and stores the conversation for analytics.
Path
Recommended path
- 1Set up the account and workspaceVerify email, choose the active workspace, and confirm the role has enough access.
- 2Create and configure the AI AgentEnter the website URL, answer setup questions, then tune behavior, fallback, domains, lead capture, and widget copy.
- 3Train and test before publishingAdd trusted sources, test in playground, then publish only after readiness checks pass.
- 4Monitor and improveUse conversations, analytics, usage, support, and admin monitoring to keep quality high.
Account and workspace setup
Create an account, verify email, accept invites, switch workspaces, and understand how workspace ownership works.