Cursor fails with 'Calling undefined' on first try

[DrewDennison]

[DrewDennison]

from: https://semgrep.slack.com/archives/C08LP9YTKN2/p1753120967095349?thread_ts=1751265006.779599&cid=C08LP9YTKN2

[manikandan-rajappan]

1. I doubt my recent cursor update (1.2.4), that was not true, because other MCP server doesn’t have this issue.
2. Tried running by building docker image & running as container OR ghcr.io/semgrep/mcp this image - all gives the same behaviour.
3. Tried running with /sse mode also - problem persist.
4. I doubt the mcp (python package) - downgraded to 1.9.4 - but that also didn’t help.
5. No clue on whether the issue is on tool registration/calling

[flaper87]

1. I doubt my recent cursor update (1.2.4), that was not true, because other MCP server doesn’t have this issue.
2. Tried running by building docker image & running as container OR ghcr.io/semgrep/mcp this image - all gives the same behaviour.
3. Tried running with /sse mode also - problem persist.
4. I doubt the mcp (python package) - downgraded to 1.9.4 - but that also didn’t help.
5. No clue on whether the issue is on tool registration/calling

Hi! Thanks for the report. Could you please share the mcp config you have in Cursor? 🙏

[manikandan-rajappan]

@flaper87
I hosted semgrep mcp on my infra, so the endpoint is internal one, even with localhost also not working.
below is the one with internal

{
  "mcpServers": {

    "semgrep": {
      "type": "streamable-http",
      "url": "http://127.0.0.1:8000/mcp"
    }
}

Belos is the cursor version details:

[flaper87]

Can you try switching to the sse endpoint?

Configuration

    "semgrep-sse": {
      "type": "streamable-http",
      "url": "http://127.0.0.1:8000/sse"
    }

MCP Server started:

Pormpt

Command

$ export SEMGREP_APP_TOKEN=....
$ uv run semgrep-mcp --transport sse 

[manikandan-rajappan]

@flaper87 even with sse, it shows the same behaviour. Attaching the video- check it on slack - https://semgrep.slack.com/archives/C08LP9YTKN2/p1753514403143439?thread_ts=1751265006.779599&cid=C08LP9YTKN2

Are you not facing this issue either with /mcp or /sse?

[flaper87]

I don't have access to this slack instance. Cna you please post the erroe here?

It'll also make it easier for future readers/contributors

[manikandan-rajappan]

@flaper87 please check it. https://www.loom.com/share/c85d78b63f70489693af856493cb25ee?sid=f92e5e8c-732a-49f9-bc3c-5bef65581fba

both /sse and /mcp behaves the same way

[flaper87]

@flaper87 please check it. https://www.loom.com/share/c85d78b63f70489693af856493cb25ee?sid=f92e5e8c-732a-49f9-bc3c-5bef65581fba

both /sse and /mcp behaves the same way

Just tried again, using the same prompt:

I'm now wondering if this is a problem with cursor itself: What model are you using?

A quick Google provided this result: https://forum.cursor.com/t/mcp-server-calling-undefined/120481

[manikandan-rajappan]

@flaper87
Thanks for sharing the discussion link.

it was in Auto mode with this models (claude-4-sonnet, gemini-2.5-pro, gpt-4o and gpt-4.1)
Tried disabling all except gemini, but that also same behaviour.
But I have one more mcp server with mindsdb - just for testing, but that tool was called instantly.

[manikandan-rajappan]

@flaper87 this behaviour is not coming on your setup? can you share the model, cursor version, let me replicate and see

[flaper87]

@flaper87 this behaviour is not coming on your setup? can you share the model, cursor version, let me replicate and see

I can't seem to replicate this 😢

These are the three configs I have tested: