During refactoring it happens all the time but happens very frequently during regular development as well.
It’s highly annoying having to babysit the agent.
During refactoring it happens all the time but happens very frequently during regular development as well.
It’s highly annoying having to babysit the agent.
Yeah. We need increase it.
Remember to click the blue VOTE button below the topic headline!
That’d be nice but getting a consistent working agent that’s stable would be nice for more than a week.
+1 it would be good if it was configurable (with an upper-limit if required).
I’d like to see this or a more stable agent. I burned a lot of tool calls today when the agent was trying to run my test suite like 3 different ways because it didn’t know what it was doing.
I complained about this and their response was that I should use Claude 3.7 MAX INSTEAD
Good reminder, the current UX doesn’t make it stand out that much so easy for some people to miss it.
Yeah it’s annoying when you run into issues what ever they might be. Hopefully both issues can be fixed
Yeah that would be great. Yeah that happens some times especially during a refactor it can go a bit hay-wire even if you reset the context by opening a new chat.
Haha what that’s hilarious. I think there’s a need for more than 25 for regular pro usage as well. MAX can do 200 currently right? But that will be quite expensive.
I think it should be increased too. If it can’t be increased more than 25, an option should be added to the settings and after the 25 tool call limit, it should automatically charge 1 credit for every 25 call limit. For example, if the transaction ends with 53 tool calls. It should charge 3 credits in total but not wait. Waiting is waste of time.
To still have a bit control it would be good to set a limit similar to Usage Based setting, so perhaps a 50 or so tool call limit so it doesnt go into the sky. And to use the buzzer on completion. (That one may need to be louder but it does notify me when its done)
Yeah something like that would be necessary I think. See this bug report I made for the sound volume being too low Notification Sound Too Low tl;dr they might fix it soon.
Hey all!
This limit was put in place to avoid the agent getting itself into a loop, and using up a ton of tool calls not doing a lot, but you can always click continue if you hit the 25 tool call limit, and it’ll carry on from where it was (and use another premium request!).
Thanks for the background, I figured it was due to something like that, primarily a cost issue was what I thought.
For me the instances where I run into the limit it’s clear it’s not in a loop.
And it happens frequently enough that it feels like it kills my productivity.
Would it be possible maybe to set up a custom limit and having it continue automatically until the limit is reached ?
Currently hitting it quite often and that’s never a loop issue too as Frulfump said
I’d bet we’re a lot in that case
A cost issue is very understandable, and I’m perfectly happy with it using up more requests.
Will feed this back to the team, we may be at a point where the risk of a loop is low enough to allow people to opt-in to higher limits!
That’s great to hear, thanks!