Unable to automatically summarize; the summarization feature cannot specify a model

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

When using a custom API:

  1. Context cannot be automatically summarized;(How should this problem be solved?)

  2. The summarization feature does not go through the custom API model but still deducts account fees.

Steps to Reproduce

/Summarize Or automatic summarization

Operating System

MacOS

Version Information

Version: 3.1.0-pre.12.patch.0 (Universal)
VSCode Version: 1.105.1
Commit: 157ffe1c6dac02720a7c5d875229565305625f10
Date: 2026-04-06T16:39:06.552Z
Layout: editor
Build Type: Stable
Release Track: Nightly
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Darwin x64 22.4.0

Does this stop you from using Cursor

No - Cursor works, but with this issue

5

Cursor doesn’t send credentials when summarizing.

If we try to send credentials (patch workbench.desktop.main.js), it still works on version 2.x.
But in version 3.x, it’s almost completely blocked by Cursor on the server side.

It seems like Cursor is trying to create bugs in the custom API to force users to use their API.

In my opinion, this bug is quite simple and they could fix it in a few minutes. But they let this situation persist for several months:

  • Client side: sending credentials in the summarize request. (They used to send them, but they were removed.)
  • Server side: forwarding requests to the custom model if credentials are available (instead of blocking them).

Screenshot 2026-04-07 at 22.02.00
Screenshot 2026-04-07 at 22.02.002586×1320 499 KB

9

Hey, this is a known issue. Summarization currently doesn’t use credentials from a custom API, it goes through Cursor’s internal infrastructure. If your subscription isn’t active, the request just fails.

Similar reports are already up:

The team is aware. There’s no ETA yet, but reporting it helps with prioritization.

Current workaround is to start a new chat before the context fills up, around 60 to 70% usage.

Let me know if you have any questions.