Documentation

Everything needed to build an assistant on your own content and put it in front of customers — on your website or through the API.

Quickstart API keys

Getting started

Getting startedCreate a workspace, build a grounded chatbot, and put it in front of customers.Core conceptsWorkspaces, roles, channels and the boundaries the platform enforces.

Building

ChatbotsInstructions, model settings, knowledge and the publish flow.AgentsTools, memory, structured output and human approval.WorkflowsTyped nodes, validation before running, and a per-step run history.KnowledgeHow collections group your sources into retrievable, cited answers.

Operating

ConversationsOne inbox for every channel, with summaries and human replies.CRMContacts, stages, tags, notes and activity, linked to conversations.AnalyticsConversations, messages, resolution rate and token spend over time.IntegrationsOutgoing webhooks, Slack and SMTP, with credentials encrypted at rest.

Deploying

Website embedPut the assistant on your own site with a script tag and an allowed-domain list.API authenticationCreate a workspace API key and authenticate a server-to-server request.Chat APISend a message to a chatbot and stream the reply.ErrorsOne error envelope, stable codes, and what to do about each.Rate limitsThe ceilings applied to the public API and the widget.

Reference

ExamplesWorking clients for the streaming chat API.