Skip to main content
In a few minutes, your own code will be answering questions over your team’s entire knowledge base — no retrieval pipeline, no embeddings, no infrastructure to run. This guide takes you from zero to that first answer with plain curl. For authentication and HTTP semantics, see Introduction.
Prefer not to write code? Your AI coding agent can do all of this for you — connect it to Chief and just ask. See Use Chief with your AI coding agent. This page is the manual, do-it-yourself path.

Prerequisites

Export variables for the examples:

Start a chat

POST /v1/chats starts an async workflow and returns ids immediately:
Example response (202 Accepted):

Poll for the response

Poll until response appears on the message:
Use exponential backoff between polls and cap total wait time in production.

Continue the conversation

What you just built

That’s the whole loop: your code asked a question, Chief answered it from your project’s knowledge, and you continued the thread — with no retrieval pipeline or infrastructure to run. That same primitive is the foundation for a Q&A feature in your app, a scheduled digest, an agent that reads your project, and more. See What you can build for where to take it.

Next steps

  • What you can build — concrete outcomes built on this loop
  • Chats guide — scope, intelligence modes, and listing
  • Assets guide — upload files for use in scope
  • Tools — skip curl with the Go SDK, the chief CLI, or the MCP server
  • API reference → Chats → Start a chat — full request schema