73 comments

[ 4.5 ms ] story [ 84.6 ms ] thread
I was really not expecting Apple to jump on this bandwagon, but I guess this was inevitable.
Thanks Apple, but "agentic coding" was already very possible without Xcode supporting it natively. Always gotta get your OKRs, I guess.
I wish they put their energy elsewhere like fix bugs, make faster.
My annoyance is that it sounds like I can't just use Claude Code directly in XCode? I like how Zed does it, it's not perfect, but it works really nicely.
I am already using Claude in Xcode 26.2. What did they change / add specifically in 26.3? It's not super clear behind the marketing haze.
I wonder how much of the recent Apple OS releases were done with "agentic coding".
MCP support is the real story here Means you're not locked into Claude or Codex Can plug in whatever agent you want
Was surprised they use MCP for this and not ACP.
> plug in whatever agent you want

Noob question: Does that include local models?

Unfortunately the MCP for external agent tools is flawed at the moment, it returns a format differing from its stated schema. So it doesn't work with OpenCode.
(comment deleted)
Building castles in the sky while the foundation is rotting away :/ Xcode really needs a couple of years of pure bugfix and optimization releases instead of hype-chasing.
customers pay for features, not bugfixes
As long as it's purely opt-in and before opting in no data is ever sent to some server and no source code can be changed by it, I'm okay with it.
Does agents.md allow for automatic discovery of mcp tools (Tools: run ./tool-discovery.sh)
I'm looking forward to trying the SwiftUI preview integration, though from my experience using the xcodebuildmcp and axe tools to let agents run simulators and capture screenshots, expectations will be low. It seemed like the models were capable of identifying issues like "button that should be there is not displayed", but not identifying when the layout is wrong or some element is too big.
I built an entire iOS app without opening Xcode UI even once. Why so many iOS engineers prefer XCode?
how did you debug it? The CLI tools are close to useless. I never even found a way to get console output from a device without Xcode.
First time I tried it, claude built all the files in the wrong directory lol. It's working fine now.
So far I find OpenAI’s Codex app to be the right approach for me. I can’t stand AI integrated IDE’s, it creeps me out when code starts changing at a phase that I can’t follow.

Yesterday in few hours I released an update for my mac App that I haven’t been working on for over a year. The update easily performed as expected, did a few small manual touches on the UI and the app just got approved on AppStore(like minutes ago)[0].

This is very good because normally I would not remember much about the code, so doing an update for a long forgotten code becomes huge pain.

Good for Apple but I think I feel most comfortable on Codex app. I think I like having the AI separated from the IDE so I feel in control in the IDE.

[0] Codex implemented the functionality demo on the paywall, if you want to see it: https://apps.apple.com/us/app/crystalclear-sound-switcher/id...

How about adding a horizontal scroll to sidebars? No?

"Agentic this", "agentic that"... It's literally just an LLM in a while() loop with some exposed tools.

I have not been able to switch to Opus 4.5 in XCode. It defaults to Sonnet 4.5 and I couldn't find where to change it (or if it's possible). Anyone know?
[flagged]
This is huge news. Human-in-the-loop development is essential for actual software velocity gains. The current tooling around agent enabled iOS dev leaves a lot to be desired. Every time I work on web-dev tasks I'm jealous of the tooling.
I don't think I'm ready for my phone apps to get even more sloppy...

I wonder if they used this internally to write iOS 26? Would explain some things...