WebFetch results are not persisted in transcript

[aono_shang]

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Hey folks, ran into something while building a custom command that does a lot of web scraping.
I have a command that uses WebFetch to grab multiple API documentation pages. The problem is - WebFetch results don’t seem to be saved in the conversation transcript at all.
What happens:
Agent fetches a URL with WebFetch, gets the content
If the conversation gets long or something interrupts, that fetched content is just… gone
Agent ends up re-fetching the same URLs over and over because it can’t “remember” what it already got
I checked the transcript file and there’s no trace of the WebFetch response content
My workaround was to explicitly tell the agent to write every WebFetch result to a file immediately (like a cache), and to do it serially (one URL at a time, fetch → write → next). Otherwise parallel fetches mean multiple results could be lost if anything goes wrong mid-way.
Is this expected behavior? Seems like other tool results (file reads, grep, etc.) do get saved to transcript. Would be nice if WebFetch results were persisted too, or at least have some way to cache them automatically.

Steps to Reproduce

1. Create a simple command or prompt that tells the agent to fetch multiple URLs using WebFetch (e.g., “Fetch these 5 documentation pages and summarize them”)

2. Let the agent run and fetch a few URLs successfully

3. Either:

   • Wait for the conversation to get long enough that context gets summarized, OR
   • Interrupt/stop the agent mid-way and resume

4. Ask the agent to reference or use the previously fetched content

5. Observe: The agent has no memory of the WebFetch results and will attempt to re-fetch the same URLs

6. (Optional) Check the transcript file at ~/.cursor/projects/.../agent-transcripts/xxx.txt - you’ll see [Tool call] WebFetch entries but no corresponding content in [Tool result] sections, unlike other tools like Read or Grep which do have their results saved

Screenshots / Screen Recordings

Operating System

MacOS

Current Cursor Version (Menu → About Cursor → Copy)

Version: 2.4.7
VSCode Version: 1.105.1
Commit: ca0f9bf806f235ea014a22712cbcbf5e88ca77e0
Date: 2026-01-20T20:52:38.077Z
Build Type: Stable
Release Track: Early Access
Electron: 39.2.7
Chromium: 142.0.7444.235
Node.js: 22.21.1
V8: 14.2.231.21-electron.0
OS: Darwin arm64 23.4.0

For AI issues: which model did you use?

opus 4.5 thinking

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

[deanrie]

Hey, thanks for the report and screenshots.

This looks like a bug. Other tool results (Read, Grep) are saved in the transcript, but WebFetch isn’t. Your workaround of writing it to files is a good solution for now.

I’ll pass this to the team.