Hi Cursor team,
I’m not a deeply technical user, but I use Cursor daily on a real setup: a platform repo (shared libraries + sample app) and a separate customer host repo that consumes those libraries via NuGet packages.
I tried the Agents Window with enthusiasm. In theory it should help: two repos, package updates, migrations, and agent work across both. In practice, I went back to the classic editor and I’ll stay there until Agents Window catches up.
What I expected
One place where I can:
Run agents across both repos (platform + customer host)
Run the web app (F5 / debug) and open the login page in the browser
Review changes line by line and accept/reject
Use the same reliable agent + terminal workflow I already have in classic mode
That’s the full loop: agent changes code → build → run → test in browser → approve.
What actually happened (Windows)
Agents Window felt less capable, not more
I hit errors; agents often didn’t run properly from the Agents Window
I could not run/debug my ASP.NET project from there the way I do with F5 in classic mode
I had to switch back to classic IDE to run the app locally and test in the browser
Classic mode agents work; Agents Window agents often don’t — I’ve seen reports about workspace trust / shell tools not activating in the Glass layout on Windows
So Agents Window today adds orchestration ideas but removes execution power. For my workflow, that’s backwards.
Why parity matters (concrete daily workflow)
A typical day:
Fix something in shared library/UI code (platform repo)
Build + tests
Bump package version + update customer host dependency
Run the site locally and verify login, grids, permissions
Steps 3–4 require classic IDE today (F5, debugger, terminal). Agents Window doesn’t complete the loop.
Parallel agents across two repos sound great — but if I still need classic mode to run and verify, Agents Window is an extra window, not an upgrade.
What should move from classic → Agents Window
Please prioritize parity, not just new agent features:
F5 / Run / Debug for local web apps (.NET and others)
Reliable shell + agent execution on Windows (including workspace trust in Glass layout)
Full diff review (per-hunk accept/reject like editor mode)
IntelliSense / LSP when I open files from agent results
Extensions I already rely on in classic mode
Plan mode with the same workflow as editor
Your docs say both interfaces will coexist. That’s fine — but classic editor power should be brought into Agents Window, not kept as a separate “go back if you need to actually work” escape hatch.
Suggested direction
Think: agent orchestration + full IDE in one surface, not two half-products: Agents + multi-repo + F5/run/debug + diff review + stable tools (Windows)
Until then, users like me will keep classic mode as the real workspace and treat Agents Window as experimental.
Thanks for keeping classic mode — please merge its strengths into Agents Window rather than asking us to choose.
— Windows user, multi-repo .NET (platform + customer host via NuGet)