Background Agent stuck on loading with small prompt

Where does the bug appear (feature/product)?

Background Agent (GitHub, Slack, Web, Linear)

Describe the Bug

When launching a background agent with a very small prompt, the agent remains in the “starting” state and never starts running. This started happening a few hours ago.

Steps to Reproduce

  1. Create a background agent with a very small prompt.
  2. Start the agent.
  3. Observe that the agent remains in the “starting” state indefinitely and never starts running.

Expected Behavior

The background agent should start normally and process small prompts quickly instead of remaining in the starting state indefinitely.

Operating System

MacOS

Current Cursor Version (Menu → About Cursor → Copy)

Version: 2.0.43
VSCode Version: 1.99.3
Commit: 8e4da76ad196925accaa169efcae28c45454cce0
Date: 2025-10-30T18:49:27.589Z
Electron: 34.5.8
Chromium: 132.0.6834.210
Node.js: 20.19.1
V8: 13.2.152.41-electron.0
OS: Darwin arm64 23.3.0

For AI issues: which model did you use?

GPT5

For AI issues: add Request ID with privacy disabled

bc-a6e268ae-f865-49fa-accb-3cdb72c4640b

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

Please help us Cursor :sob:

Hey, that’s certainly odd that so many agent runs are stuck, but you can try the following to narrow down the issue:

  • Check how many background agents are pending, vs completed ones. I see around 30 pending runs, which can affect new agents.
  • Open the project in a new location (or create a new project) and try running a small background agent.
    • If it works in the new location, the problem is likely specific to the current project.

That said, do you remember how so many agent runs were created in the first place?

(post deleted by author)

  • Every single one of our Cloud Agents created via the Background Agents API are getting stuck in pending. We did not change anything on our end to cause them to start getting stuck. None of the cloud agents are complete. Every one becomes pending.
  • Worktrees and local agents run fine.
  • All of these agent runs were indeed created legitimately. It does not appear that agent runs are disappearing. They’re all going to pending.
  • It fails even when creating the Cloud Agent from Cursor app instead of the API endpoint
  • It always gets to the step of “Environment ready” → “Planning next moves” and then gets stuck. It will always title the agent before getting stuck.