Skip to main content

Overview

Agent Studio is where you build AI agents and put them to work.

An agent is a model with instructions, tools, and access to your data. You give it a job ("answer support tickets", "summarise pull requests", "schedule customer follow-ups") and it does the job — calling tools, asking for information, and producing a result.

An app is several agents wired together. One agent receives the request, decides what's needed, and delegates to specialist agents. The user sees one chat; under the hood there's a small team.

You can use Agent Studio to:

  • Build a single agent in minutes — pick a model, write the instructions, attach tools.
  • Compose agents into apps with an entry-point agent that routes to specialists.
  • Trigger agents on a schedule, on a webhook, or from a chat.
  • Share an agent or app as a public chat URL.
  • Connect tools via MCP or any external API.
  • Watch what your agents do — every interaction, every tool call, every result.

If you've used a chat-style AI tool, you've used the consumer side of this. Agent Studio is the builder side: where you design the agent, hook it to your systems, and ship it.

The rest of this section walks through it, starting with Getting Started.