We’re introducing Cursor 3. It is simpler, more powerful, and built for a world where all code is written by agents, while keeping the depth of a development environment.

With the new Cursor, you can run as many agents as you want, everywhere you want: locally, in a worktree, on remote ssh, and in the cloud. And it has the best parts of the editor available when you need them.

The new interface is available as a separate window that complements the IDE. Update Cursor to try it.

We recently launched Composer 2, a frontier model with high limits. Then, with cloud, we gave agents their own computers so they can work truly autonomously. And now with Cursor 3, we’re releasing a new interface to collaborate with agents on software.

Awesome 🤠

I'm proud to announce that I designed this. Here is my design process:

we've spent the last week testing it @every! read our full vibe check:

cloud is so sick. will try today. lets go sualeh!

Never going back. I spent thousands of dollars in tokens every month. Then I realized the same amount of work I was doing (just working on three simple serverless websites), on Claude Max Plan, is just $200/mo. Never been billed extra than that. Cursor wastes tokens like crazy. Do NOT use.

'built for a world where all code is written by agents' is a hell of a positioning statement. not 'AI-assisted coding' anymore. agent-first, human-supervised. the IDE becomes an orchestration layer.

I will still not use it in production😂

the industry was NOT ready for this ecosystem, i am already in love

Cursor is really changing the world now I think im leaving the Codex app for this Now I will spend 100% of my time in Cursor, instead of 80% 😅

cursor 3 is cool but agents writing code isn't the hard part anymore. reviewing 50 PRs/hour is. the bottleneck moved from generation to judgment. who verifies the agents?

“all code will be written by agents” isn’t the wild part this is: you’re slowly becoming a reviewer, not a writer Cursor 3 just makes it obvious you prompt → it builds → you fix edge cases that’s the new dev loop

cursor this cursor that, okay bro

Looks great! I think what would really boost productivity and our ability to work with agents is if you make a mobile or web app that allows us to interact at the same level as this desktop app Everyone's building it out, and hopefully we see it from yourselves soon

Local <--> remote ftw; thank you!

Cursor already destroys every other harness when it comes to features. Just please, please, please work on bug fixes and performance optimisations. The app is literally unusable 30% of the time. Opus 4.6 max thinking tokens is no longer streaming, app sign significantly slows down unless I frequently click save, agents randomly becoming one TPS post compaction to name a few. The app just feels vibe coded. Please spend a month just polishing it.

LFG Cursors is back!

reinstalling cursor now

I was always a @windsurf fan, but today I am going to subscribe for Cursor.

@FabParad the interface problem was always underrated. spinning up agents is easy. knowing what they’re all doing at the same time is not. Cursor 3 is the first thing that actually feels built around that constraint

Interesting... Cursor is basically turning into an agent orchestration layer on top of a dev environment. 👍

I had a feeling something significant was about to drop after the silence!

@Scobleizer How is it different from claude code web/desktop?

actually pretty disappointed, there’s nothing ground breaking, seems to me like everything was present there before

@Scobleizer Orca is way better is open source and MIT licensed

@jbfja @sualehasif996 looks so happy

Sick!!! Uninstalling conductor now 😂

cursor seems like a really good harness, can i run it on my remote gpu without having constant connection on my laptop, something like tmux for terminal

Feels like the shift from writing code to orchestrating agents is becoming real. The tools are getting out of the way, now it’s about how well you can direct the system.

the real question is what happens when all code is written by agents — do we still need IDEs or just supervision terminals?

the design decision that matters: keeping the IDE and the agent interface separate. most tools are trying to collapse the two into one surface. cursor is betting that the review loop and the generation loop need to stay distinct. if a dev is reviewing agent output in the same view where the agent is generating, the feedback signal is compromised. separate windows is an opinion about where the human fits in the workflow.

bro I can't even keep up y'all are shipping so fast! Thank you!

A strong step toward the future of development. Balancing agent-based coding with deep environment control is a compelling vision, excited to explore its capabilities.