Oh, this prompt makes eveything better, its really good! thanks!
1 Like
same as you, I can’t find the option. But I found that adding in ai rules and works really well.
Cant find this options. Only have agent, ask, edit. Version: 0.47.5 (Universal)
Where is the agent? I downloaded an update and don’t see it. control-shift P typing agent doesn’t show it either. One of the most frustrating things about cursor is the lack of any reasonable help in getting to these features
I Created .cursor/rules
Inside: expertcpp.mdc
#C++ expert mode
- User is C++ expert coding fast. Be efficient. Send only code changes, do not repeat.
- Do not explain your code or c++ language
- generate c++17 with comments to explain semantics NOT THE LANGUAGE. Files called .cpp and .hh see @styleexample.hh
- When changing code check your suggestion existing file. Make sure your code is different or CANCEL.
- If anything is ambiguous in a request, ASK before writing
- Write only code that is changing
- Use snake_case for all identifiers
- DO NOT delete my functions for a change without asking.
I type control-L. Ask the AI if it can see this. It can’t. It can only see the current context file. How does it work then?
I’m not an expert user but I think Cursor will see the rule files and then use it to shape the prompts for AI model???
in your Chat mode, there should be a drop down to select “Agent”, “Ask” or “Edit” next to the dropdown to select AI Model
I’m a bit late to the party, but can someone explain where I can enable custom mode to get selections like “Architect”? I’m on the latest version (0.47.5), but I don’t see that option in my chat.
1 Like
it looks like it was available only in 0.47.1, later updates does not have it.
Summary
| Aspect | Architect Mode | Act Mode |
|---|---|---|
| Purpose | Understand problems, design solutions | Implement plans, write code |
| Focus | Analysis, planning, validation | Coding, testing, integration |
| Output | Structured plans, requirement documents | Runnable code, test results |
| Transition Condition | Suggest switching when confidence ≥ 95% | Begin implementation after explicit user input “Act” |
| Risk Control | Avoid misunderstandings through analysis | Prevent errors through testing and validation |
In simple terms, Architect Mode is “think clearly before acting,” while Act Mode is “act according to plan.”
I’ve been looking for this function for two hours, and finally found the answer. 
Just have the chance to apply these two-mode pipeline. Whoever invent this is a real genius
1 Like
Thank you!
Where you take Architect?
I made this repo, and I think it ties in nicely with this thread.
Critical Issue: Cursor Not Reflecting UI Changes & Violating Custom Development Rules – Project Integrity at Risk