Agent can't web search properly

Web search is completely broken in the cursor CLI as well from what I can tell. In my case, the desktop app works fine but the CLI does not. I am lobbying my company hard to let us use a different AI coding tool which prioritizes the CLI client more.

I reported this here.

Still the same issue. Web search completely inexistent.

1 Like

This has been going on for the entire month AT LEAST for now. How is this possible? Can we please get some reaction here?

1 Like

Linking similar threads here for visibility. This is an issue many users are experiencing and it should be prioritized.

Can this bug be prioritized and fixed in the next sprint? It’s been present for 3 months.

They don’t care. @condor doesn’t even answer my questions about it.

How could a simple web search tool be unavailable till now? This bug should be prioritized and fixed ASAP.

1 Like

This is critical

1 Like

Request ID

  • c671c3fc-4d10-4270-a43a-45291b00e7a6

Check the latest response if it could help accelerate pinpointing the issue.

It’s ridiculous to see the model failing to do a web search, and fallbacks to generate a Python script to query and parse the raw HTML like this:

python - <<'PY'
import re, urllib.request
html=urllib.request.urlopen('https://piclist.cn/en/configure',timeout=30).read().decode('utf-8','ignore')
# find the h3 that contains WebDAV
m=re.search(r'<h3[^>]*>[\s\S]*?WebDAV[\s\S]*?</h3>', html, flags=re.I)
print('found', bool(m))
if not m:
    raise SystemExit
start=m.start(); end=m.end()
print('h3 tag snippet:', re.sub(r'\s+',' ', html[start:end])[:300])
chunk=html[end:end+12000]
# cut at next h3
m2=re.search(r'<h3[^>]*>', chunk, flags=re.I)
if m2:
    chunk=chunk[:m2.start()]
chunk_no_code=re.sub(r'<pre[\s\S]*?</pre>','\n[CODE]\n',chunk)
text=re.sub(r'<[^>]+>','\n',chunk_no_code)
text=re.sub(r'\n{2,}','\n',text)
lines=[l.strip() for l in text.split('\n') if l.strip() and l.strip()!='[CODE]']
print('--- text ---')
print('\n'.join(lines[:160]))
PY

In addition, you can see that the local Browser Tab is also not working. The model could not manage to input a field, click a button or even scroll inside the Browser Tab.

@Colin could you take a look at the thread?

Any updates? Now it even returns completely irrelevant web search results. Web search tool is completely unusable.

I’ll switch to Claude Code if this critical feature could not be fixed soon. It’s just intolerable.

Request ID

e0d18e40-ab2c-4c40-be83-a271b8a1f202

Screenshot

Hey all!

Thanks to @Artemonim, I was able to dig into at least one issue with web search and filed a bug with the team.

@Hakula, yours looks like the same issue. The previous message you sent to the agent is being used as the search term, not the one shown in the UI. Your search results are irrelevant on the surface, but they kind of make sense if you consider the actual search term was zsh (994-1022).

Sorry about this one! It’s in the team’s hands now, but I’ll keep an eye on it.

1 Like

Hey, I hate to do this but is there any update on this issue?

I believe my issue in the following thread is related

FYI, I’ve switched to Brave Search MCP to replace the built-in web search tool.

Partially fixed in Nightly

  • Agents can run searches and get correct results
  • Subagents have the issues described in the initial bug report
Version: 2.4.0-pre.29.patch.0 (system setup)
VSCode Version: 1.105.1
Commit: 31ba4eea39481b2e13065083d98cf92613175e00
Date: 2026-01-13T20:27:33.722Z
Build Type: Stable
Release Track: Nightly
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Windows_NT x64 10.0.22631