Message Too Long on auto

Support Bug Reports
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Tried to use Agent review and fix the issues it found. The agent is new, the model is Auto, doesn’t have any context. IDE button pasted text into the input box and added 6 file links. I’ve hit Send, and was greeted with the result Message Too Long, Your message is too long. Please try again with a shorter message and fewer/smaller attached items.

So I’ve removed file links from the field and hit Send again. Still, Message too long. The text itself is:

Verify these issues exist and fix them:

Bug 1:
The `documentation-classifier.mdc` file contains the entire content of itself repeated twice without proper separation (line 134 onwards), creating duplicated sections. Similarly, `documentation-organization.mdc` and `documentation-quick-decision.mdc` both have their content duplicated at the end. This appears to be a copy-paste error during file creation. The duplicate content should be removed to keep these cursor rules clean and maintainable.

Bug 2:
Real S3 access credentials (`<actual data>`) and OpenAI proxy secrets (`<actual data>`) are exposed in configuration documentation files. These appear to be actual credentials used in the Yandex Cloud S3 and should never be committed to version control. These should be replaced with placeholder values like `your-access-key` or `your-secret-key` before deployment.

So I’ve created a new Agent, made sure no file is open (so no current file in context), no MCP is used, and copied the text in there. Still Message Too Long.

Steps to Reproduce

  • Be on Auto
  • Use Agent Review
  • Try to fix all found problems
  • Be greeted with Message Too Long error

Expected Behavior

  • either the Agent starts working on the request
  • or it details what happened (how much context is used, what is the max length, what is added to the context) so user can fix it

Screenshots / Screen Recordings

2025-11-23_12-58-08.png
2025-11-23_12-58-08.png966×464 72.3 KB

Operating System

Linux

Current Cursor Version (Menu → About Cursor → Copy)

Version: 2.1.20
VSCode Version: 1.105.1
Commit: a8d8905b06c8da1739af6f789efd59c28ac2a680
Date: 2025-11-22T14:51:59.472Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Linux x64 6.17.4-76061704-generic

For AI issues: which model did you use?

Auto

For AI issues: add Request ID with privacy disabled

21489ae6-f61a-4253-a390-b3e0c5bae387

Additional Information

It seems to be tied to the amount of open tabs. After closing all open tabs, the new agent was able to start working. I think you should check which tabs were opened or activated lately and include only those, not all of them.

Does this stop you from using Cursor

No - Cursor works, but with this issue

2

Hey, thanks for the report. You’re right that closing all open tabs fixes the issue, and this confirms that Agent Review is mistakenly including all open tabs in the context calculation.

Your suggestion makes sense: Agent Review should probably only use recently active tabs instead of all of them, since that can quickly hit context limits.

I’ll pass this on to the team. For now, the workaround of closing unused tabs before running Agent Review is the best option.

Thanks again for such a thorough investigation.

4

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