Hacker News Reader

Comment by timcobb | Hacker News Reader

How much does it cost to run Claude Code 24 hrs/day like this. Does the $200/month plan hold up? My spend on Cursor has been high... I'm wondering if I can just collapse it into a 200/month CC subscription.

View on HN

This guy tested it: https://she-llac.com/claude-limits

"Suspiciously precise floats, or, how I got Claude's real limits" 19hs ago 25 points https://news.ycombinator.com/item?id=46756742

OTOH, with ChatGPT/Codex limits are less of a problem, in general.

Because Codex effectively rate limits you by being so slow.

It’s slower but generally spits out more reliable code, IMHO.

Not in my experience recently. Are you using it via Cursor, or how?

If you're using it 24h/day you probably will run into it unless you're very careful about managing context and/or the requests are punctuated by long-running tool use (e.g. time-consuming test suites).

I'm on the $200/month plan, and I do have Claude running unattended for hours at a time. I have hit the weekly limits at times of particularly aggressive use (multiple sessions in parallel for hours at a time) but since it's involved more than one session at the time, I'm not really sure how close I got to the equivalent of one session 24/7.

How do you prompt it so it can run many hours at a time? Or do you run it in some kind of loop that you manage yourself?

if you do enough planning up front, you can get a swarm of agents to run for hours on end completing all the tasks autonomously. I have a test project that uses github issues as a kanban board, I iterate with the primary chat interface to refine a local ROADMAP.md file and then tell it "get started"

it took several sessions of this to refine the workflow docs to something claude + subagents would stick to regarding branching strategy and integration requirements, but it runs well enough. my main bottleneck now is CI, but I still hit the weekly limit on claude max from just a handful of these sessions each week, and it's about all the spare time I have for manual QA anyway

Make it write a plan or todo list, and then make it spawn sub agents to execute. If you have the main agent do the work it will soon go off plan and stop, but when it's just spawning agents, it will be willing to run for a very long time.

Also take care to tell it what it should solve itself rather than stop and ask you for help with, and run it contained so you can turn on yolo mode.

There's a daily token limit. While I've never run into that limit while operating Claude as a human, I have received warnings that I'm getting close. I imagine that an unattended setup will blow through the token limit in not too much time.

I built a similar autonomous loop using LangGraph for a publishing backend and the raw API costs were significantly higher than $200. The subscription model likely has opaque usage limits that trigger fairly quickly under that kind of load. For a bootstrapped setup I usually find the predictability of the API bill worth the premium over hitting a black box limit.

I have no first-hand experience with the Max subscription (which the $200 plan is) but having read a few discussions here and on GitHub [1] it seems that Anthropic has tanked the usage limits in the last few weeks and thus I would argue that you would run into limits pretty quick if you using it (unsupervised) for 24h each day.

1) https://github.com/anthropics/claude-code/issues/16157

The employee in that thread claims that they didn't change the rate limits and when they look into it, it's usually noob error.

It's a really low quality github issue thread. People making claims with zero data, just vibes, yet it's trivial to get the data to back the claims.

The guy who responds to the employee even claims that his "lawyer is already on the case" in some lame threat.

I wonder how many of these people had 30 MCP servers installed using 150k of their 200k context in every prompt.

Yea there are some weird replies in that thread. My few highlights were "This is my livelihood, not a hobby or sideproject" or "I just purchased a third $200 MAX plan and instantly hit rate limits". While I agree that it might not be Anthropics fault I've gotta admit that I found Anthropic to be rather vague regarding their rate limits. They seem to have totally dynamic rate limits based on usage and not a fixed "messages per hour" or "tokens per hour" based approach. Their free tier usage page states "Also, the number of messages you can send will vary based on demand, and we may impose other types of usage limits to ensure fair access to all users." [1] while the Pro plan page just says "During peak hours, the Pro plan offers at least five times the usage per session compared to our free service." [2] and Max then 5x or 20x it depending on the price you pay. If they just have more demand or reduced the free tier rate limit, all plans have a reduced limit and it will be totally within their communication. OpenAI at least gives you a specific amount of messages per timeframe (which I find more transparent). [4]

1) https://support.claude.com/en/articles/8602283-about-free-cl... 2) https://support.claude.com/en/articles/8324991-about-claude-... 3) https://support.claude.com/en/articles/11014257-about-claude... 4) https://help.openai.com/en/articles/11909943-gpt-52-in-chatg...

Reply on news.ycombinator.com