Getting started

This platform builds AI assistants on your own content and gives you two ways to reach customers: a script tag on your website, or an HTTP API you call from your own application. This page takes you from a new account to a published chatbot.

Create a workspace#

Sign up at /sign-up. The first account creates an organization and a workspace, and becomes its owner. A workspace is the unit of isolation: chatbots, knowledge, conversations and contacts all belong to exactly one, and nothing crosses between them.

Build a chatbot#

1. 1

   Create it

   Chatbots → New chatbot. Give it a name; everything else has a working default.

2. 2

   Write the instructions

   The Instructions tab holds the system prompt: who the assistant is, what it should do, and what it should refuse. Keep it specific; this is the highest-leverage setting on the page.

3. 3

   Ground it

   Create a collection under Knowledge, add a document or a URL, wait for it to reach Ready, then attach it on the chatbot's Knowledge tab. Answers will cite what they used.

4. 4

   Test it

   The Playground runs against the configuration you just saved, streaming the reply exactly as a visitor would see it. Conversations from here are recorded on the playground channel so they never mix with real traffic.

5. 5

   Publish it

   On Deploy, add the domains allowed to use the widget, then activate the chatbot. It answers nothing until it is active.

Choose an integration#

Both paths reach the same chatbot, the same knowledge and the same streaming response.

Where to go next#

• Core concepts — workspaces, roles, channels and how isolation works.
• Knowledge — collections, and how retrieval-augmented generation works here.
• API authentication — create a key and make your first call.