Transcripts no longer exported in full

Support Bug Reports
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

As of a few days ago, using Export Transcript in the chat window produces summaries of conversations, not complete copies. My chat transcripts have gone from being complete multi-megabyte files to a few kilobytes, even after long conversations.

It appears that Cursor is now actively excluding a number of critical categories of information from transcripts, including:

  • Actual content of files edited and changed
  • Thinking processes
  • Terminal input and output
  • Git commit statements

For example, a section of transcript that would have usually included these detailed changes now only says something like:

"I can see the issue. The text on lines (line number range) and (line number) claims (feature) support that doesn’t exist yet. Let me update this to reflect the current reality while noting that (feature) support is coming soon.

Done. Let me verify the changes look correct.

Changes applied."

Transcripts should be just that, transcripts. If I see a line of text in the chat window, it should be captured in the transcript file. Looking at my history, this behavior seems to have started after January 16, 2026.

Steps to Reproduce

  • Conduct a conversation with a model (e.g. Opus 4.5), especially one with thinking, terminal and code generation
  • Export Transcript
  • View output

Expected Behavior

Incomplete export of transcript (summaries, not actual transcripts)

Operating System

Linux

Version Information

Environment details:

Version: 2.4.21
VSCode Version: 1.105.1
Commit: dc8361355d709f306d5159635a677a571b277bc0
Date: 2026-01-22T16:57:59.675Z
Build Type: Stable
Release Track: Default
Electron: 39.2.7
Chromium: 142.0.7444.235
Node.js: 22.21.1
V8: 14.2.231.21-electron.0
OS: Linux x64 6.8.0-90-generic

For AI issues: which model did you use?

Opus 4.5 (thinking)

Does this stop you from using Cursor

No - Cursor works, but with this issue

3

Hey, thanks for the report, I can see the issue. Transcript export should really keep everything you can see in the chat (file changes, thinking, terminal output), not just summaries.

This looks like a regression since mid-January. I’ve passed it to the team, your report helps us prioritize it.

To help with the investigation, can you share:

  • The Request ID from the affected chat (top-right chat menu > Copy Request ID)
  • If possible, an example exported file where we can see the missing content

Let me know if you find any workaround.

5

Hi Dean, thanks for the prompt response. A request ID of an example chat would be 470549c7-891f-4a99-af30-1adeda1c013a . Is there a way that I can DM you the associated transcript for this chat?

8

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

When I try to export the chat transcript, the commands the agent ran (and the input/output) from those commands aren’t exported with it. See the following screenshot of the exported transcript on the left (which doesn’t have the commands the agent ran) and the actual chat history on the right (which you can see the collapsible command input/output).

Steps to Reproduce

If you open a chat and ask it to run commands and then export the transcript, the command input/output does not appear.

Screenshots / Screen Recordings

Screenshot 2026-03-25 at 04.10.39.png
Screenshot 2026-03-25 at 04.10.39.png2462Ă—1202 555 KB

Operating System

MacOS

Version Information

Version: 2.6.20 (Universal)
VSCode Version: 1.105.1
Commit: b29eb4ee5f9f6d1cb2afbc09070198d3ea6ad760
Date: 2026-03-17T01:50:02.404Z (1 wk ago)
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Darwin arm64 25.3.0

Does this stop you from using Cursor

No - Cursor works, but with this issue

10

Can we please get an update on this regression? There is a ton of missing stuff in the chat export, including the agent thinking as well.

11

Hey, thanks for the reminder, and sorry for going quiet.

Honest update: the original ticket for this regression was closed without a fix. After looking into it, the current behavior is that thinking blocks and collapsed items (terminal input/output, tool calls) are intentionally not included in Export Transcript, because technically they aren’t chat messages. Expanding export to include these blocks is being tracked as a feature request, not a bug fix. There’s no ETA.

If on the latest version you’re seeing something missing besides thinking and terminal (for example, actual text like file changes, git commits, or agent replies), that’s a different issue. Please open a new thread with your Cursor version, an export example, and a screenshot of what’s in the chat, and we’ll take a fresh look.

cc @rsb

12

Thanks for the update, @deanrie . Beyond the regression, I think I recall you (or one of your colleagues) mentioning in a different thread that certain long-chain thinking outputs are inhibited from being logged by model providers?

13

Hey, yep, that’s correct. For extended thinking models, providers intentionally only return a summary of the reasoning, not the full chain-of-thought. Anthropic clearly documents this for Claude 4 models: the Messages API returns a summary of the full thinking process, not the raw reasoning. OpenAI behaves similarly with reasoning models.

What you see in chat as the thinking block is that same summary we get from the provider. We don’t get the full chain ourselves, so there’s nothing extra to include in the transcript export beyond that.

Separate issue: right now, collapsed thinking blocks (the summary you see in the UI) aren’t included in Export Transcript. That’s on our side and it’s tracked as a feature request, like I mentioned above.

14

it is very hard for us to get the feedback we need to improve our tools unless there is a way to export everything possible, including permitted thinking raw tool input and output.