Composer / Auto failing to connect when using images

Support Bug Reports
, , ,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

When uploading an image (useful for debugging errors) the agent fails to connect.

Steps to Reproduce

Have only Composer 2.0 and 2.5 available. You can have auto or either of the specific models selected.

Upload an image.

Expected Behavior

Image is processed and routed to the model

Operating System

MacOS

Version Information

Version: 2.6.14
VSCode Version: 1.105.1
Commit: eb1c4e0702d201d1226d2a7afb25c501c2e56080
Date: 2026-03-08T15:36:54.709Z
Build Type: Stable
Release Track: Default
Electron: 39.6.0
Chromium: 142.0.7444.265
Node.js: 22.22.0
V8: 14.2.231.22-electron.0
OS: Darwin arm64 25.3.0

For AI issues: which model did you use?

Auto, Composer 2 and 2.5

For AI issues: add Request ID with privacy disabled

Request ID: 636dbbe8-7273-4082-9e42-310a669d19e5
{“error”:“ERROR_PROVIDER_ERROR”,“details”:{“title”:“Provider Error”,“detail”:“We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.”,“isRetryable”:false,“additionalInfo”:{},“buttons”:,“planChoices”:},“isExpected”:true}
Provider Error We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.
IAi: Provider Error We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.
at o5A (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34244:23670)
at r5A (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34244:22658)
at m5A (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34245:6285)
at kou.run (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34245:10400)
at async _Oa.runAgentLoop (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:46808:10024)
at async N3u.streamFromAgentBackend (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:46862:9277)
at async N3u.getAgentStreamResponse (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:46862:13663)
at async bMe.submitChatMaybeAbortCurrent (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34305:17597)
at async Object.Ea [as onSubmit] (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:45787:4826)
at async vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:45761:56460

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

8

This is a known bug. Our team is aware and working on it.

Workaround for now:

  • Start a new chat without images if you need Composer 2/2.5

  • If you need to work with images, select a multimodal model (like Claude or GPT) from the model picker instead of Auto or Composer

You may also find this related thread helpful: Cursor having issues connecting to composer 2?

Also, your Cursor version (2.6.14, from March 2026) is quite outdated. While this specific bug is server-side, updating to the latest version will include many other fixes and improvements.

Our team is aware and investigating.

9

Do you have an ETA on when Composer will work with PDFs? Our org has disabled other models for now bricking my workflows and forcing me to consider migrating to another Coding agent/IDE

10

PDFs use the same mechanism as images. Composer 2/2.5 can’t natively process multimodal content (images or PDFs), so it relies on another model to handle that. Since your org has restricted the available models, those requests fail.

The quickest unblock: ask your admin to enable at least one multimodal-capable model. That would let Composer hand off image and PDF processing while keeping Composer as the primary agent model.

We don’t have a specific ETA for native multimodal support in Composer. I understand this is a significant workflow disruption, especially with limited model availability. If your admin has questions about which models to enable, they can reach out to your account team for guidance.