# Ideas for improving Cursor AI🌟 **URL:** \ **Category:** Feature Requests\ **Created:** [June 9, 2025, 2:55pm UTC](https://forum.cursor.com/t/ideas-for-improving-cursor-ai/102057 "2025-06-09T14:55:51Z")\ **Posts on this page:** 6\ **Page:** 1
**Author:** ![wtester](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/wtester/32/64278_2.png) [@wtester](https://forum.cursor.com/u/wtester)\ **Post date:** [June 9, 2025, 2:55pm UTC](https://forum.cursor.com/t/ideas-for-improving-cursor-ai/102057/1 "2025-06-09T14:55:51Z")
1. Many users ask to increase the 25 tool call limit, but there’s another approach: add an option for the agent to use multiple premium requests if it exceeds 25 tool calls. For example, we can set a limit of 2–3 premium requests for the agent. If it goes over the 25-call limit, it automatically uses a second premium request, and so on. But if it doesn’t exceed 25 calls, it only uses one premium request. 2. Real-time hints for the agent. While working, the agent might take a long time to think or drift off-topic. So we could provide guiding hints during task execution, or introduce a pause feature — clicking it would allow us to enter a hint manually. 3. Collaborative mode with two models acting as agents. One handles planning, and the other executes the task. If a problem arises during execution, the process switches to the planning model to rethink the approach. This way, we could spend two premium requests — one for reasoning, and one for execution — and they would switch between each other, assisting one another dynamically. * * * ---
**Author:** ![iperezr](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/iperezr/32/62383_2.png) [@iperezr](https://forum.cursor.com/u/iperezr)\ **Post date:** [June 10, 2025, 12:55am UTC](https://forum.cursor.com/t/ideas-for-improving-cursor-ai/102057/2 "2025-06-10T00:55:29Z")
Really like 1. I’ve found myself wishing to do 2 many times, unclear on how practical it is. ---
**Author:** ![duxyz](https://avatars.discourse-cdn.com/v4/letter/d/258eb7/32.png) [@duxyz](https://forum.cursor.com/u/duxyz)\ **Post date:** [June 10, 2025, 1:49am UTC](https://forum.cursor.com/t/ideas-for-improving-cursor-ai/102057/3 "2025-06-10T01:49:05Z")
These are all great ideas! 👌 I’ve run into that 25 call limit so many times, especially when things get a bit more technical. The agent just stops after a while and I have to type “continue”, which breaks the flow. Being able to use multiple premium requests automatically would make a big difference. The second idea is also very useful. There have been plenty of moments where I wished I could pause the agent and give it a quick hint or add more info mid-task. That kind of real-time guidance would save a lot of back-and-forth. And the third one? Love it. I’m using Taskmaster for something similar but if Cursor had that built in, it would be fantastic. Thanks for sharing these. Cursor devs, please pick this up! ---
**Author:** ![matgenehr](https://avatars.discourse-cdn.com/v4/letter/m/f04885/32.png) [@matgenehr](https://forum.cursor.com/u/matgenehr)\ **Post date:** [June 10, 2025, 2:56am UTC](https://forum.cursor.com/t/ideas-for-improving-cursor-ai/102057/4 "2025-06-10T02:56:11Z")
Great ideas. ---
**Author:** ![friken0](https://sea3.discourse-cdn.com/cursor1/user_avatar/forum.cursor.com/friken0/32/62544_2.png) [@friken0](https://forum.cursor.com/u/friken0)\ **Post date:** [June 10, 2025, 9:30am UTC](https://forum.cursor.com/t/ideas-for-improving-cursor-ai/102057/5 "2025-06-10T09:30:48Z")
I like the idea of pay-more for higher toolcall limits per prompt. down for that as long as my ai-buddy doesnt whoops down a rabbit hole while I’m at lunch not watching and run up a bill that makes long-distance carrier calls of the 1980s reasonable. ---
**Author:** ![leoing](https://avatars.discourse-cdn.com/v4/letter/l/dec6dc/32.png) [@leoing](https://forum.cursor.com/u/leoing)\ **Post date:** [June 10, 2025, 3:18pm UTC](https://forum.cursor.com/t/ideas-for-improving-cursor-ai/102057/6 "2025-06-10T15:18:38Z")
One could set a personal token limit, each worth 25 tool calls. But I guess it’s against the pricing strategy. I think base price $20 is to lure us into maxxing. Which is a very generous offer, to be clear.