[BUG] Claude Code (Pro) session limit reached after 20 minutes

LOL - on the Pro plan used to work for a couple of hours until hitting the 5 hour limit say an hour or give or take from the limit ... A couple of days back got kicked out in under an hour I recall!!!!! Have been using another product since - will have to cancel my subscription if this is not corrected - but guess that this is the company's intent as bcherry in another thread stressed that current performance is quite in line as advertised(sic!) and had the thread closed!!! Pretty clear messaging I believe ....

This has been happening to me for the last few days. The sessions seem to be getting shorter and shorter. I have also been kicked in the middle of the system making significant changes leaving me with hours of work trying to clean up the mess. I went ahead and cancelled my service this morning and will use alternatives until there is some kind of comment as to what is going on here.

Noticing the CC Pro plan w/ Sonnet 4.5 is hitting the Session Limit Reached throttle after less than an hour of boilerplate work. Not even continuous work. More like 3-5 min mind-numbing jobs (fixing compiler warnings) a few times in less than an hour.

C'mon Anthropic - we're using Sonnet 4.5 on the Pro plan (yearly subscriber) and we can barely get an hour of CC usage. There needs to be better visibility into these usage limits. Either that or just give up the mirage of a "Pro" plan and force everyone on the "Max" plan b/c that sure seems like where you're pushing us.

Be transparent; this passive-aggressive rate limiting is nonsense.

I’m also experiencing this issue. After starting a session, I consistently hit the “session limit reached” message after about 20 minutes of usage. This happens even when I’m not running heavy workloads. It seems like the session timeout is shorter than expected, and it interrupts ongoing work.

Had the same thing happen today. Session limit hit after 4 Messages sent to Claude. Never happend to me before. See picture attached: Session limit reached after 11 Minutes?

<img width="1134" height="104" alt="Image" src="https://github.com/user-attachments/assets/ab01845f-61d5-428b-aa81-c0d93ded1b37" />

I subscribe to the Pro plan ($20.00/month). 35 minutes later I hit session limits trying to resolve an Eclipse/Maven issue.

Only 45 minutes after subscribing, I CANCELED my subscription and continue to use Grok.

Ditto. Usage has become comically limited regardless of complexity or volume. half the time is spent trying to work with or around its flawed 'memory' to continue from a prior conversation. Then abrupt session death. Completely different and unacceptable experience on Pro plan compared to the past few months. Ugh.

Same

Also experiencing this

I've been using Claude Haiku, Sonnet and Opus in CoPilot for months and I always had a good experience. Yesterday I decided I'd sign up to Claude Pro and give Claude Code a try. I hit the session limit within an hour or so (I didn't even know it was a thing). This morning, the same thing happened, I hit the limit in 1 hour. I'm not impressed. I pay a lot less for CoPilot and get a lot more.

This issue has been inactive for 30 days. If the issue is still occurring, please comment to let us know. Otherwise, this issue will be automatically closed in 30 days for housekeeping purposes.

Still happening for me, I hit the limit comically fast on Opus while troubleshooting a script made by Claude...
I subscribed for the Pro plan today

Still happening for me, I hit the limit comically fast on Opus while troubleshooting a script made by Claude... I subscribed for the Pro plan today

My recommendation would be to only use Opus when working on complex problems. Otherwise you'll burn through your tokens. Still with Sonnet. Or even Haiku on dead simple work such as scripts.

Still happening for me, I hit the limit comically fast on Opus while troubleshooting a script made by Claude... I subscribed for the Pro plan today

same here. Opus is activate on default which is really insanely fast depleting my tokens. I switched to Sonnet now which is way better in means of token consumption.

Switched to Claude today to try it out. The first three prompts hallucinated that it made changes to the file system and didn't actually do anything. The next prompt failed halfway through, from hitting session limit

I literally got one 99 line file before hitting my limit? This if $20 worth of product? Copilot on the same Opus model ran me dozens of prompts and thousands of lines of code.

Same issue here, I wonder if it's intended behaviour? In a single conversation, I barely got past the planning stage and started the implementation when I got hit with the limit already, in ~20 minutes. In comparison, I can work non-stop for about 4 hours at least in Google Antigravity with their entry level plan. Using Sonnet 4.6 and manually approving each eddit, no fully automatic editing or using Opus.

Same issue. Burned through the 5h limit in 45 minutes using Sonnet 4.6
Unfortunately it's completely useless. I've seen it was reported several times since october 2025, no response.

Over a few days I managed to get better by delegating simple tasks to Haiku, and for more complex ones, plan with Opus and implement with Sonnet. Also using compact feature a lot, and starting new conversations whenever possible rather than reusing old ones. Hope this helps others a bit.

But we should not go through all this trouble just to be able to use this thing 2-3 hours instead of 30 minutes in a day.

Edit: I take this back sadly, I managed to get a bit more out of Claude on a daily basis, but now I'm hit with a weekly limit on Wednesday morning. I think the only solution is to cancel and subscribe for another service?

Closing for now — inactive for too long. Please open a new issue if this is still relevant.

The frustration of Claude Pro (and Max) users over steadily shrinking usage limits is well-documented and has been intensifying in recent months. Developers on Reddit and Anthropic's official Discord channel have repeatedly reported that their token allocations are depleting far faster than before, with some claiming that criticism has been silenced by Discord moderators. The Register One user estimated, based on a token-level analysis of Claude Code logs, that there had been roughly a 60% reduction in token usage limits. The Register The situation worsened significantly after New Year 2026, when Opus 4.5 usage limits were drastically reduced compared to all previous weeks since the model's launch — one user consumed 27% of their weekly allocation after just a single day of normal use. GitHub Another wave of complaints arrived around February 23–25, 2026, with users noticing dramatically faster limit depletion coinciding exactly with the Claude 4.6 rollout, as paying Pro subscribers hit rate limits mid-week with no explanation from Anthropic. GitHub The entire AI industry is moving toward explicit access rationing — both Anthropic and Google now publish hard limits where vague "limited access" language once stood, and Anthropic has even blocked third-party tools that were bypassing flat-rate subscription limits. PYMNTS Anthropic itself partially acknowledged the issue by launching a temporary promotion in March 2026 that doubles limits during off-peak hours, which arguably confirms that standard limits are insufficient for many users. PYMNTS The experience matches a pattern reported by hundreds of users: for the same subscription price, you're getting progressively less actual output.

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.