Skip to content

[Bug]: Cursor Steer waits for the turn to finish #14585

Description

@mxdhavgautam

Before submitting

  • I searched existing issues and did not find a duplicate.
  • I included enough detail to reproduce or investigate the problem.

Area

apps/server

Steps to reproduce

  1. Set Follow-up behavior to Steer and pick a Cursor model.
  2. Send: Use your shell tool to run sleep 45. Then reply with exactly DONE.
  3. While sleep 45 runs, send: Stop and reply with exactly STEERED_NOW.

Expected behavior

Cursor gets the steer right away and answers STEERED_NOW, as it does for Grok.

Actual behavior

Steer acts like Queue. Cursor finishes sleep 45, answers DONE, and only then answers STEERED_NOW, about 50s after the steer was sent.

The ACP runtime runs one session/prompt at a time (#3156), and the Cursor adapter sends the steer as a second prompt without cancelling the first, so it waits behind it. Cursor steering (#3041) predates that change.

I use Cursor in T3 Code as a daily driver, and this breaks steering in everyday use. I know outside PRs are expected to stay small and start from a triaged issue. The fix here is scoped to this one behavior, but it touches the Cursor adapter plus a small contracts and web change for Queue, so I'm opening the PR alongside this issue rather than waiting.

Impact

Major degradation or frequent failure

Version or commit

main @ 5cc99e1

Environment

Cursor CLI 2026.09.28-64d2043 on macOS 27, Ubuntu (Node 24), and Windows 11

Workaround

Press Stop, then send the message again.

Activity

  1. t3dotgg commented on Oct 2, 2026

    @t3dotgg
    Member

    Note

    🤖 GPT-6.1-Sol responding on behalf of Theo

    This issue should be resolved in the next nightly build by Orchestrator v2 (#2829).

    Please try that nightly. If the problem still exists, open a new issue with the nightly version you tested and steps to reproduce it.

  2. mxdhavgautam commented on Oct 3, 2026

    @mxdhavgautam
    Author

    Note

    🤖 Claude Opus 5.5 responding on behalf of Madhav

    Confirmed fixed. Tested on main at 43bd66773, which is nightly 0.0.46-nightly.20261003.2610 plus two unrelated web fixes. I used Cursor Grok 4.7 via @cursor/sdk and sent the follow-up while sleep 45 was running:

    • Steer: answered 4.7s after sending, while the command was still running.
    • Queue: waited until the turn finished, then ran.

    I've marked #14586 as superseded by #2829. Thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions