I’ve been running into tons of issues with Cursor’s built-in terminal / CMD handling. Here are some examples:
-
Monorepo test structure bug
I have my tests inbackend/testsinside a monorepo.
Cursor automatically doescd backend— fine — but then it locks the working directory. After that, it keeps trying tocd backendagain and again, which fails and loops endlessly. -
When Cursor quits the terminal
When Cursor closes or resets the terminal, it sometimes writes a strayqinto the input buffer.
Then the next command (likeuv run pytest ...) ends up asquv run pytestand fails. -
Terminal output visibility
The response/output is often hidden — I can’t see what actually happened.
The “Open in Terminal” option isn’t great either; I’d much rather see the output inline, right in the chat window or embedded terminal area (like an accordion view). -
Slow to reflect completion
When a command finishes, it often takes ages for Cursor to reflect that it’s done.
Sometimes it looks completely stuck — I’m never sure if it’s hanging or just lagging.
I’ve ended up manually copying and pasting failed test output into the chat window just to keep moving, even though the terminal was spawned by the agent itself.
At this point, it’s nearly unusable for running terminal commands — especially in monorepos where the folder structure matters a lot.
Anyone else experiencing this? Any known workarounds or fixes?