New in Claude Code: /ultraplan Claude builds an implementation plan for you on the web. You can read it and edit it, then run the plan on the web or back in your terminal. Available now in preview for all users with CC on the web enabled.

nice! can we get your comments on this though

no but let me know if you have any questions about ultraplan 👨‍💼

Ultraplan uses roughly the same number of tokens (and subscription rate limits) as plan mode. See the docs for more:

A few questions on plan vs ultraplan. I think the insight here is that implementation can sometimes require a local environment/interactivity, but planning can happen in the cloud since it's mostly about reading code & understanding intent.

community is going to become burnt out & fatigued on daily new features when current features (and nerfing) or degradation in 4.6 1M (esp when context is > 50%) is obvious. if yall could go a week without shipping new features and making the old ones better than would be great

the "ultra" prefix is very misleading if it's just plan mode but with a different UI

ULTRAPLANIAN MODE: ENABLED

usage limits are shit now. I'm easily hitting the 5-hour session limit even with a single Claude Code thread and taking plenty of short breaks... I doubt I will be renewing. I'm on Max ($100/month).

@lydiahallie This is such a peculiar flow, I’m curious why go to the web and break out of terminal? Seems jarring. just for plan formatting?

I don't want any more features. Fix basic shit.

for those wondering how this can be used... first idea that comes to mind is that it avoids the situation of Claude keeping your terminal hostage during a planning phase you prompt → planning happens remotely (cloud) → you keep working → then review → then execute

do you guys stop shipping daily after world domination or when?😂

It's been really interesting to see the things you've added to Claude Code over the last two months. You're basically codifying what my process has been for the last year. Would you be interested in talking about this more in some way? Maybe an SF?

What is the point of releasing new features when you’ve nerfed the core model? At least give us the option to choose if we want to burn tokens. Not admitting or announcing the change is some coward shit on yalls part

Is this your way of substituting for the decreased thinking tokens? Fuck your efficiency gains bring back my thinking !

Ultra... Attempted to use it, it 'failed' in the middle and from there I told it to continue, and instead of finishing plan, started building feature. Looks like more and more feature you ship are full of bugs, and this is getting anoying.

I first thought this was a new $1000 plan

"interview me in detail using the AskUserQuestionTool about literally anything: technical implementation, UI & UX, concerns, tradeoffs, etc. but make sure the questions are not obvious be very in-depth and continue interviewing me continually until it's complete, then write the spec to the file" I use it often for in-depth planning mode, might be a good addition to something that has ultra in it

It was there for a week, but very unstable. 30 min limit makes no sense because I was going to use ultraplan on the go on the way to office.

I switched to Codex yesterday because, honestly, the current usage limits here are just not working for me. I get that one person leaving doesn’t matter in the short term, but this does add up over time. When people start realizing they’re more productive with something like Codex on a $100 plan compared to Claude’s limits, the choice becomes obvious. At the end of the day, developers care about one thing: being able to work without constantly worrying about limits. Right now, the restrictions make it hard to stay productive, and that’s a big problem. You might not feel the impact immediately, but long term, this kind of friction pushes users away. I’ll probably come back when either the usage limits improve or new models actually make the experience worth it again. Until then, this just isn’t it.

Worst UX ever. Terminal, then web to click on « approve plan » and teleport back to terminal If the goal is to waste time and complicate the user experience, then success!

GenUI 🤩

Nice! But where’s buddy? People are asking

Built sth similar as a Claude skill a month ago 😄 Love where you're taking it - come find me when you add mermaid diagrams, questions, custom JSX components, code annotations or image uploads 👀

1. Why present this via a web UI? Why not output a markdown file? 2. How does this compare to something like superpowers? Is the goal to replace some of those third party services which have multi-step planning approaches?

Can you fix how dumb opus 4.6 has been since Mythos leaked instead of shipping new features taht will not work as well with dumb opus?

When to plan vs ultraplan?

@Pranit was right about everything

What’s the real added benefit to local planning? If I’m waiting for the plan anyways for the implementation? Is it mostly a UI nicety? Or are there additional performance benefits?

K but when do we get Mythos pease thanks :3

Been running agents in production for months. The biggest failure mode is never the code - it's the plan. Agents without clear scope spiral into 40-minute refactor loops. /ultraplan making planning a first-class step is exactly right. That's where 80% of agent quality is decided.

lol plannotator @backnotprop

I honestly thought this was a $500 pricing plan

Claude code on web / cloud works very poorly in my experience. It’s unbelievable slow. And many times it just gets stuck and eventually times out after using up a full session usage.

Cool feature! One issue I keep running into — /ultraplan always resolves to a cached directory on the cloud instead of my actual repo workspace. The plan agent ends up with no project context. Looks like others are hitting this too (issue #44815). Hoping for a fix soon!