i have been working in product for 10 years, been a head of product, and a founder of multiple companies, and i would literally join cursor’s product team for 6 months just to help them figure out what to do with this. i love their product but it’s crucial to get this right
I’m with you, I’m a senior solution architect with a large consulting firm, they should think about engaging some people. They got the funding. But I think its a case of not knowing what they don’t know, because they don’t have the right people and culture in place. Point in fact, if cursor actually had a product manager, they would literally live on this forum. 
Instead, we have community ambassadors and community engineers, who I’m betting are multi-hatting other things as well..
I’m baffled as to why they would do this. Cursor is built on VS Code, arguably the most versatile IDE on the market. The one size fits all solution they’re trying to push is insanely dumb. The out of the box planning tool sucks, and doesn’t do what I want it to do. You’re a company full of software engineers, how is it lost on you that everyone has different needs and workflows? I’m not going to experiment with prompts just so I can circumvent whatever shadow instructions you’ve injected into the planning mode, then append my own massive prompt just to accomplish something I was able to do yesterday. This is a massive time waste.
Way to drop the ball guys. I’m finding another tool that won’t rug pull essential features.
Showing up here again to beat the drum for custom modes. Please bring it back!
I just found out (had a break in coding for a couple of weeks for family issues) and I am really angry. Considering canceling my MAX plan right now and here. How could they take this away from me? I had a couple of fine-tuned custom modes I used everyday. Now I would not even attempt to run a larger change!
Bumping here to remind the crowd at large!! UpVote this!!
I would like to break this down from a UX perspective.
Some key facts:
-
Switching to a “new” agent is the primary method of managing the context window. Cursor staff have explicitly demonstrated this workflow in at least one or more videos.
-
Adding commands back into the new agent every time you spawn it, is extra user input that can be error prone, subject to un-repeatability, and runs counter to automation patterns.
-
SOME form of agent customization is the most friction-free and obvious pattern to address this.
-
Placing this functionality in the “modes” dropdown is not scalable, also requires an extra user input, and runs counter to UX/UI conventions around agent customization patterns we see across the industry.
Proposal:
Consider allowing user to access custom agents from the agent sidebar. This is a more scalable approach, and is a cleaner UX pattern.
An ALTERNATIVE, but less desirable approach, would be to at least let users clear the context window for a spawned agent. This would somewhat alleviate the number of times a user is forced to set up a new agent for a task, and take a bit of pressure off of the missing custom agents feature.
2026 Jan 11 and still no custom modes… at this point, I don’t know if this will ever be brought back
Update: I guess it kinda makes sense now with the modes module they have made, the agent + plan + debug + ask, its complicated to have custom modes with these since using the agent mode can automatically pick plan mode and so on, but still, I believe there is a workaround with that, like, at least allow the existing agent mode have the ability to edit its tools and that would be fine tbh
bumping this again.. please bring back custom modes pleeeeaaaassseeee somebody heaaarrr meeeeeeee
Well… rest in peace to custom modes and memories, it was good run while it lasted, now we are facing another unfair change, GPT-5.4 being max mode only for both 1M and 270K
Good afternoon and firstly I’d say thanks for an amazing product!
I’d kindly ask to take my +1 vote for custom agent modes (or add basic ASK + MCP mode).
Point is that we are actively using MCP for debug, research and it makes us turn on Agent mode. But each time we do research we MUST write “do not edit anything” etc.
Of course we can use rules and commands instead, but it’s an LLM-way of protection, which is known to be non-deterministic. We’d be happy to have deterministic protection of write-operations with enabled MCP servers (it could be as a custom mode or +1 base mode like Ask + MCP).
Best regards!
Please bring back custom modes!!
This should be a feature, an AI tool should have several customizable workflow options without being restrictive plus people migrating from competitors are going to expect it. Which is exactly me, I am new to cursor and I expected this feature to be there
Adding my voice to the choir here. Missing this feature may be a deal breaker for us, possibly the thing that determines if Cursor will be adopted by our entire team or not. The UX experience of having a custom agent is way way way more intuitive for the average non-engineer than having to remember a set of error-prone slash commands.
Please bring custom modes back as well as the ability to specify different default models for the current/existing built-in modes as well, like it was before.