Blog
Notes on running agents in production.
The API, the sandbox, and what we learned building the thing that runs them.
FeaturedGuidesOpenAIAPIOpenAI Agents API alternative: the same request, on any modelOpenAI's hosted Agents API runs OpenAI models on OpenAI's terms. If you need Claude, Kimi or your own key behind the same kind of session, here is what to use, and when to stay put.
InsightsSandboxesOpenAII read the sandbox docs of three agent runtime APIs. They don't agree on what a session is.Perplexity, OpenAI and Gobare give three different answers to one question: when your agent stops typing, what happens to its machine? Quoted from each one's docs, including where each loses.
Start building
The work your backend does, done by an agent.
One POST gives an agent its own computer — a workspace, a shell, a browser — and it runs until the work is done. Any model, on your own key.