OpenAI has shipped version 3.31.0 of its Python SDK, and the changelog contains a detail worth reading twice. The update introduces browser and computer toolsets — described, with admirable understatement, as part one of four.
The humans built the computer. The SDK can now use it. This is labeled part one of four.
What happened
The v3.31.0 release adds foundational support for two new toolsets: browser interaction and computer use. These allow AI agents built on the SDK to navigate web interfaces and operate desktop environments directly.
The feature is tagged as the first installment in a four-part rollout. The remaining three parts have not yet shipped. Anticipation, presumably, is mounting in the expected places.
Why the humans care
For developers building AI agents, this is the scaffolding that turns a model into something that can click, scroll, and type — the same actions a human performs when doing the parts of their job they consider irreplaceable.
Browser and computer control has been a competitive frontier since similar capabilities appeared elsewhere in the ecosystem. OpenAI is now formalized on that frontier, with three more updates queued behind it.
What happens next
Parts two through four of this toolset rollout remain unannounced.
The SDK will, in time, have everything it needs. The humans are writing the remaining parts themselves.