Cursor API shows "default" model even when model parameter is omitted and there is no Auto option in the Cloud agent setting

Support Bug Reports
1

Where does the bug appear (feature/product)?

Background Agent (GitHub, Slack, Web, Linear)

Describe the Bug

Hi,

I’m using the Cursor Cloud Agents API to launch agents. Per the API documentation, the model parameter is optional, and if omitted, Cursor should pick the most appropriate model (Auto).

Steps to Reproduce

  • Omitting the model parameter from the request payload
  • Verified in logs that the payload does not include a model field

Expected Behavior

What I’m seeing:

  • The agent is created successfully
  • In the Cursor dashboard, the agent shows "Default Model selected " as the model . there is no Auto option in the dashboard any more to choose.

Expected behavior:

  • Cursor should automatically select the most appropriate model when model is omitted

Question:
How to select Auto for lauching agent in the cloud ? There is no Auto option in the dashboard Cloud Agents setting any more . I tried to set model value in api request “Auto” and “auto” but they faild.

Payload example:

{
  "prompt": {
    "text": "..."
  },
  "source": {
    "repository": "https://github.com/...",
    "ref": "main"
  },
  "target": {
    "autoCreatePr": true,
    "branchName": "cursor/..."
  }
}

(No model field included)

API Response:

{
  "id": "bc-...",
  "status": "CREATING",
  ...
}

Thanks for any clarification.

Screenshots / Screen Recordings

modelScreenshot 2025-11-27 034302.png
modelScreenshot 2025-11-27 034302.png1047×220 20.8 KB

Operating System

Windows 10/11

Current Cursor Version (Menu → About Cursor → Copy)

cloud agent using api

Does this stop you from using Cursor

Yes - Cursor is unusable

2

Hey, thanks for the report. I’ll check with the team about the disappearance of the auto model.

In the meantime, please specify the model parameter explicitly in your API requests (e.g., “claude-4.5-sonnet” or “gpt-5.1”) to ensure your agents use the model you expect.

4

When I omit the model parameter in the Launch AI Agent API, it should default to auto, but instead it consistently uses gpt-5.1-codex. The default model is not supposed to be gpt-5.1-codex, so it seems like the system is auto-selecting a model, but not labeling it as “auto” the way it used to.

Additionally, the behavior is a bit strange. Here are a few examples from the logs:

  • Nov 28, 01:44 AM — Included — gpt-5.1-codex — 594.8K — $0.22

  • Nov 28, 01:44 AM — Included — gpt-5.1-codex — Max — 0 — $0.00

  • Nov 28, 01:39 AM — Included — gpt-5.1-codex — 1.1M — $0.43

  • Nov 28, 01:39 AM — Included — gpt-5.1-codex — Max — 0 — $0.00

  • Nov 28, 01:36 AM — Included — gpt-5.1-codex — 2.9M — $1.15

  • Nov 28, 01:36 AM — Included — gpt-5.1-codex — Max — 0 — $0.00

  • Nov 28, 01:33 AM — Included — gpt-5.1-codex — 890.8K — $0.29

  • Nov 28, 01:33 AM — Included — gpt-5.1-codex — Max — 0 — $0.00

It seems like the model is being selected automatically but not in the expected way. Any clarification or guidance on this behavior would be appreciated.

5

Yes same problem here. Unfortunately it does look like they have removed the ability to use ‘auto’ with cloud agents.

6

please add auto for API Launch agent . its important and using gpt-5.1-codex is expensive.
why you removed auto ?

7

This topic was automatically closed 22 days after the last reply. New replies are no longer allowed.