Agent has trouble detecting available MCP tools

Support Bug Reports
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

When prompting the agent to utilize an MCP tool, it has a hard time figuring out how to use the tool. I gave it a prompt like this:

@link-to-honeybadger-fault
@Branch 

Please review this PR's changes. You can use the Honeybadger MCP server to look up details on the provided fault, and determine if these changes are a good fix.

The agent proceeds to look all over the place to try and find the MCP tool to use. This took about five minutes before it finally figured it out. (see screenshot)

Steps to Reproduce

Give the AI agent a link to a Honeybadger fault and instruct it to use the Honeybadger MCP server integration to fetch information about it

Expected Behavior

The agent should know what MCP servers and tools it has available without having to do an exhaustive search in weird places.

Screenshots / Screen Recordings

Screenshot 2025-12-16 at 6.33.28 PM.png
Screenshot 2025-12-16 at 6.33.28 PM.png2382Ă—1328 241 KB

Operating System

MacOS

Current Cursor Version (Menu → About Cursor → Copy)

Version: 2.2.20 (Universal)
VSCode Version: 1.105.1
Commit: b3573281c4775bfc6bba466bf6563d3d498d1070
Date: 2025-12-12T06:29:26.017Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Darwin arm64 25.1.0

For AI issues: which model did you use?

GPT 5.1 Codex

Does this stop you from using Cursor

No - Cursor works, but with this issue

Can get MCP tools working in Cursor anymore
2

Hey, thanks for the report.

This is a known issue: the Agent sometimes doesn’t detect available MCP tools right away, even when the server is connected correctly. There are a few similar reports (for example, with Playwright and HuggingFace MCP).

For deeper analysis, could you share:

  1. The Request ID from the problematic interaction (chat context menu > Copy Request ID)
  2. MCP server logs from the Output panel (View > Output > select the MCP server from the dropdown)

This will help engineers understand why the Agent performs an “exhaustive search” instead of using the tool directly.

1 Like
3

Thanks for your quick response!

Here’s what I see in the MCP Output logs from the time frame:

2025-12-16 21:48:49.094 [info] Handling CreateClient action
2025-12-16 21:48:49.094 [info] CreateClient completed, server stored: true
2025-12-16 21:48:49.099 [info] Handling ListOfferings action, server stored: true
2025-12-16 21:48:49.099 [info] Connected to stdio server, fetching offerings
2025-12-16 21:48:49.130 [info] listOfferings: Found 13 tools
2025-12-16 21:48:49.130 [info] Found 13 tools, 0 prompts, and 0 resources
2025-12-16 21:48:49.495 [info] Handling CreateClient action
2025-12-16 21:48:49.495 [info] CreateClient completed, server stored: true
2025-12-16 21:48:49.500 [info] Handling ListOfferings action, server stored: true
2025-12-16 21:48:49.500 [info] Connected to stdio server, fetching offerings
2025-12-16 21:48:49.504 [info] listOfferings: Found 13 tools
2025-12-16 21:48:49.504 [info] Found 13 tools, 0 prompts, and 0 resources

Request ID: df7849a5-a614-4634-8025-05ae06148d1d

1 Like
5

Here’s some related posts about MCP bugs that your team needs to investigate as well:-

6

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