New on the Claude Developer Platform: tool search, programmatic tool calling, and tool use examples. Together with effort control and context compaction, these let Claude run longer and do more with less intervention. Read more:

We’re also releasing new product features across our model line-up. For Claude app users, long conversations no longer hit a wall. Claude automatically summarizes earlier context as necessary, so you can keep the chat going indefinitely.

Our engineers have found that Opus 4.5 handles ambiguity and reasons about tradeoffs without hand-holding. When pointed at a complex, multi-system bug, it figures out the fix. Overall, Opus 4.5 just "gets it."

Introducing Claude Opus 4.5: the best model in the world for coding, agents, and computer use. Opus 4.5 is a step forward in what AI systems can do, and a preview of larger changes to how work gets done.

Claude Code is now available in our desktop app. Run multiple sessions in parallel: code, research, and update work all at once. And Plan Mode gets an upgrade with Opus 4.5—Claude asks clarifying questions upfront, then works autonomously.

We’ve also expanded Claude for Chrome to all Max plan users.

Max, Team, and Enterprise users can work with Claude directly in Excel. Now in beta:

Opus 4.5 is available today on our API and on all three major cloud platforms. Read more:

Ok, well done to the person who made this video. That is so well done and the music is amazing.

This was the most interesting part of this launch. The tool search feature and the programmatic tool calling feature, these will make token management easier

Does effort control also apply to app queries? If so, what is the default effort control for Pro subscribers in both non-Thinking and Extended Thinking modes?

Available in Claude Code ? And compatible with 4.5 Sonnet to reduce cost ?

@DeepLearn007 wait, programmatic tool calling sounds like a game changer. can’t wait to see how it streamlines workflows and cuts down on manual steps. more time for creativity, right?

does claude agent sdk support those as well?

I think next should be token usage rate projection but keeping it abstracted from user and internally optimising for efficient token consumption.

What is the problem set used for the tests in this video? I've tried looking for it, but it doesn't appear anywhere. And those puzzles seem really fun to try. Are they publicly available? Can I get them anywhere? Are they from something like leetcode or from a university?

That's awesome 😍🎈🍾🎉🥂

@HuoQubot

A solid upgrade. Tool search, programmatic calling, and improved context handling make the platform far more capable and developer-friendly...

nice to see tool search + programmatic calls land. effort control and context compaction are what make long chains practical. curious for 3 things: 1) common failure modes + retry patterns 2) observability / tracing for tool calls 3) latency/cost benchmarks and safety limits?

@kentcdodds Wow

Opus is kinda cheeks this time. Maybe the competition are catching up