Image: Anthropic
Developers say Anthropic's Claude Agent SDK is using up the five hour usage limit on Pro and Max plans faster than the same work does in Claude Code. The claim got wider attention after T3 Code creator Theo Browne said he had confirmed it was a bug on Anthropic's side.
What Theo found
Browne said on X that all Agent SDK usage drains the five hour limit faster than using Claude Code directly. He said weekly limits are not affected and that he had reported the problem to Anthropic.
Spent a lot of time on this today. I confirmed it's a real bug on Anthropic's side. To be clear: this does not affect weekly limits.
— Theo - t3.gg (@theo) October 25, 2026
All Agents SDK usage will drain your 5h limits faster than using Claude Code directly.
I have reported this to Anthropic, hoping for a fix soon.
The screenshot he shared compares runs on the same account. Interactive Claude Code sessions moved the five hour meter by 7 percent each, while two T3 Code runs moved it by 13 and 16 percent. The weekly meter rose by 2 percent in every run. The summary in the image puts the gap at about 1.7 times for the same tokens. In a follow up post, Browne guessed that a recent boost to five hour limits was not applied to Agent SDK traffic. Anthropic announced that boost in September.
Who is affected
The Agent SDK lets developers build apps on top of Claude Code, and T3 Code is one of them. The issue spread after a T3 Code user said a single session on a Max plan emptied his whole five hour window in about 30 minutes.
Just downloaded T3 Code and hit 5-hour limits on my Claude Max plan for the first time. I was running just one session where I asked it to fix a few payment related bugs for a project, and it emptied the whole 5-hour limit in ~30 mins.
— DeepakNess (@DeepakNesss) October 24, 2026
Other users have echoed the 1.7 times figure, though it comes from Browne's own tests and not from Anthropic. A GitHub issue on the Claude Code repository has tracked a similar gap since September. Several users posted their own tests there, with some showing larger gaps. Browne posted his findings there too, and T3 Code is tracking the problem on its side.
Wait, what? Claude in T3 Code eats your 5-hour limit about 1.7x faster than Claude Code 😳
— Madhu (@madhugogineni97) October 25, 2026
@theo says it's a real bug on Anthropic's side that hits anything on the Agent SDK.
Some users have gone further. One post asked whether Claude's desktop app could also be affected, and said weekly limits feel lower too. Neither claim has been confirmed.
bro, the Agents SDK drains my Claude 5h limit faster than Claude Code
— James Malsawm (@JamesMalsawm) October 25, 2026
Anyone else?
No word from Anthropic yet
At the time of this writing, Anthropic had not publicly confirmed the bug or announced a fix. Anthropic staff had not replied on the GitHub issue, and the Claude status page listed no incident about usage limits. It only showed unrelated model errors.
Until there is a fix, some developers say they are running their work in Claude Code directly. This is not the first time Claude users have raised concerns about usage limits. TechcityAuthority has covered earlier questions about how Claude Max limits are applied.
Comments