Nice, only got the pro plan and didn't want to get Nickel and Dimed per usage so ended up ignoring claude code, but will give it another go now to see how far I can get with the Pro plan.
Turned out to be a blessing as it led me to discovering augmentcode.com which I'm now using. It's a monthly usage cap but has a unique context engine where it indexes your entire code base which gives its agents access to a lot more content about your code base which helps when adding features to match your existing implementations without needing to explicitly feed it context in every agent message. It still uses Claude Sonnet 4 [1] under the hood so the results are great.
Blows me away what AI agents are capable of these days, it's especially nice when they end up creating new Tailwind UIs (Vue) that's better looking than what I was even thinking about creating.
Seems like it's a "limited" version of the offering though:
> Pro ($20/month): Average users can send approximately 45 messages with Claude every 5 hours, OR send approximately 10-40 prompts with Claude Code every 5 hours
> Model access: Pro plan subscribers can access Sonnet 4, but won’t be able to use Opus 4 with Claude Code
Most importantly no Opus 4. Still a great nice to have!
I haven't hit a limit yet on the $200 plan, but did hit the limit quite frequently on the $100 plan FWIW. I can't imagine these Pro limits will be very useful except for very light stuff.
Is it still worth, I mean how's Sonnet vs Opus in terms of daily coding? I'm no asking about token usage, that's too abstract for me, but can it cover the whole codebase, and such.
For doing 'boring' day to day things in python I find them more or less the same. For doing non-trivial things in Rust, Opus is much better. Sonnet is much more prone to producing garbage that is just completely wrong in those cases. However I find Gemini 2.5 Pro better than both for harder programming problems.
Gemini is better than Claude in cursor, but in Claude coder I think Sonnet is better than Gemini (in cursor agent, can't use Gemini without the bridge or something else and I don't care about that case, may as well use aider or something similar).
I guess it depends? It seems a limited offering if you're planning to go all-in on agents, but on the other hand one of the most common complaints about Claude Code is that its costs are unpredictable and weird.
Your mileage may vary dramatically, but I don't think I would subscribe to Pro just for this. It's a "nice to have" on top of my existing Pro subscription whose services are the bulk of the value I'm paying for.
The biggest problem I have with Claude Code is that if, for some reason, the past discussion is not in the terminal then there is no way to review it. This is crucial to understand why my project went one way and not the other.
Aider might be less polished but at least it keeps an .md file with the chat history. That's why I am back to Aider + claude-opus-4.
This is an important point. I hope they'll address this soon. I've just started tinkering with Code and took it for granted that I wouldn't lose the "conversation" when the terminal restarted.
Maybe just ask it to save off the contents of the session as it goes?
I've found it useful to have Claude Code basically take notes in a Markdown file as it's working. When needed, edit this manually + other context and feed it into a new session. This has been fairly useful for longer-running work.
I've switched as soon as Jetbrains Junie was released. Before blowing $100 on cloude code was far too much. CLI was a stopgap for my tooling, but overall Junie is so much nicer to use and cheaper.
p.s. claude keeps trying to charge my card for subscription and wouldn't let me cancel it. Now on the basic plan some of the older projects aren't accessible so they've totally lost me. Online tons of horror stories people getting banned while unable to cancel their membership.
Does anyone know if this is available for the Teams plans as well? I still use a teams plan (5 users where I only use one seat). I did this before Anthropic introduced the Max plans. Would love to migrate to Max plan, but I don't have the time to recreate dozens of projects with all the attachments.
Is someone from Anthropic reading this? Is it possible to downgrade Teams plan to an individual Max plan and keep all the projects?
I have found Cline with Claude 4 to be light years better than using Claude Code (doing full vibe coding). The planning step and heavy communication of Cline leads to strong code, while Claude code pumps out disconnected spaghetti code. Is this the same experience others are having? I'd love to use the cheaper Claude code if it could perform better
Cursor still obscures Claude to me in every way, since it uses 3.5 and 4.0 Claude models, has almost no request limit from my personal uses, either. $20 goes possibly exponentially further on Cursor than Claude does for now. This could change in a week, though.
I spent several hours and a few dollars with Claude Code last night on the PAYG plan so this is a welcome change if it gives me more for my Claude subscription.
Do people actually get value out of the monthly plans? I've been using API access to both Claude and OAI and end up paying less, while I keep hearing about people being rate limited with lower usage. Am I missing something here? The only one that made sense to me was Cursor, but that comes with extra features and model switching.
Other than that, API + custom UI seems better since you're paying for the usage instead of competing with other flat-fee users for what seem to be overprovisioned resources (claude.ai was always throwing https 500 or 'sorry we're too busy' errors for me until I switched to API).
Not sure if it's cheaper though - I never checked.
It's insane how bad the first party chat apps are though. That's reason enough to use the API.
This did not when I tried it a few days ago but I also noticed it lets you use your API key after logging in through the web interface. I'll need to try again, really annoying that you can't just set $ANTHROPIC_API_KEY and have it work
This is really possible. I installed Claude code many months ago, and never used it - but may have logged in in at that time. When sonnet 4 was released I tried it again, and started using it with just my API keys (I do not have a paid subscription).
they are definitely not doing a good job on the release - that page doesn't mention Pro including it, and it's not actually available on Pro yet for half a dozen posters on this thread.
I remember needing a subscription just to authenticate with the Claude Code tool (at least when it came out) even though it spends from my API balance kinda felt like a scam.
I wonder how the pricing works out between using Opus in Claude Code pay-as-you-go vs the $100/mo plan. All I remember is easily spending $25 in a Claude Code session if I let the context stay too big. So maybe the sub isn't a bad deal.
I use APIs for coding with Aider (gemini 2.5 pro is my preferred model), but still subscribe to ChatGPT plus because being able to use better models like O3 and 4.1 in chat, plus unlimited image generation, Realtime voice/video, Deep Research and Codex (as a backup agent while I'm manually driving Aider) ends up being worth the price.
I would subscribe to Max if they had a coding agent interface on web like OpenAI's Codex, but without a novel implementation like that the monthly plan doesn't make sense since you can just use API, OpenRouter, etc.
> without a novel implementation like that the monthly plan doesn't make sense since you can just use API, OpenRouter, etc.
what a weird reply. whether the monthly plan makes sense depends on 1) whether it does useful things for you 2) whether you'd otherwise spend more than $20/month in API credits (which is low single digit hours of it churning away, in my experience).
claude code is definitely...more proactive than aider with gemini or sonnet-4. claude code will just go and do a whole bunch of stuff, asking questions of you and of itself, even compared to aider in architect mode. it does seem to get stuck less often, too.
you can just use claude code with anthropic API pricing, so I'd suggest you just try it and see if it's better for you personally or not.
I think you can now use Claude Code while only paying for Pro, ie. the chat bot web app. The API is a separate pay-as-you-go plan, or has been until now. But I'm not quite sure, they're super confusing about this.
Turned out to be a blessing as it led me to discovering augmentcode.com which I'm now using. It's a monthly usage cap but has a unique context engine where it indexes your entire code base which gives its agents access to a lot more content about your code base which helps when adding features to match your existing implementations without needing to explicitly feed it context in every agent message. It still uses Claude Sonnet 4 [1] under the hood so the results are great.
Blows me away what AI agents are capable of these days, it's especially nice when they end up creating new Tailwind UIs (Vue) that's better looking than what I was even thinking about creating.
[1] https://www.augmentcode.com/blog/claude-sonnet-4-the-best-mo...