I’v been using claude-3.7-sonnet-max in agent mode since its release and its been great, but it stopped working suddenly couple of hrs ago. When I gave it a prompt, it called the required tools as expected, but stopped completely from making the required code changes. What is going on?!
3 Likes
I also have, I write “create folders”, he writes “I’ll create it now, I’m starting” and that’s it.
+1 Can confirm it’s an issue right now. Just wanted to demonstrate the tool to my partner and the results were as described by OP.
+1 here too
Hey, at the moment it seems everything is already fixed. Let me know if you still encounter any issues.
Nothing is fixed yet.
no,I still has this problem
please tell me how to use it ,thanks ,i pay for it ,but dont work ,so ?
I’ve tried to switch to 0.46.11. It didn’t work
give up. I choose cline.
I believe paying users deserve compensation, as this failure has caused them tangible losses.（This failure persisted throughout the entire day.）
1 Like
Hey, it seems like we had some issues under the hood, unfortunately!
All appears to be resolved now, and we are looking into this internally to ensure this doesn’t happen again.
If anyone has lost a significant number of fast requests, please drop us an email at [email protected], and we’ll be happy to assist you with compensation!
1 Like
For users, if they find the product unusable after a few attempts, they are likely to stop using it altogether. Therefore, while the number of quick requests consumed may not be high, the time wasted and their perception of the product will inevitably change.
What many users likely desire is a stable and predictable experience, rather than frequent updates riddled with errors. The former is controllable, while the latter is chaotic. I sincerely recommend conducting more testing before rolling out significant changes.
“Offense wins games, defense wins championships.”
I hope you continue to grow and improve.
Everything is working as expected now. I also noticed a faster response times. 
This issue does not appear to resolved. I continue to have the agent reply as if it will follow up on my request, and then it does nothing. I am primarily trying to use it to invoke MCP tools. I have confirmed there is no issue with the tools or MCP server - the agent recognizes and suggests it will use them, but then does nothing. If I give it other general prompts, it seems to respond normally. This includes prompts that have it create files/code. So it seems like it might be a lingering issue specifically when it would/should invoke a custom MCP tool. It responds as if it will, but then does nothing. As further confirmation that it is not an issue with my MCP server/tools, this issue has been somewhat sporadic. Yesterday, I was able to get it using the MCP tools successfully when I set the model to “Autoselect” mode, but not when selecting a specific model (I tried all claude variants). Today, it is no longer working in autoselect mode either. I have not made any changes to my cursor installation in the meantime.
1 Like
This was improved yesterday, but today I’m back to pulling teeth with Cursor. My coworkers are reporting the same things.
It refuses to do anything. Or, if we’re lucky, it will make a few changes then ask something like “do you want me to continue”. Even when we tell it to continue, it often takes no action before asking another question.
Its working perfectly for me.
still have this issue the agent seems to not actually make edits to the file