Skip to main content


I’m trying to use Figma MCP 

But keeps getting server error 500 when using get_image

Final speech.

 

You need to have desktop mode install and enable the Enable Dev Mode MCP Server function.

 

 

 


Already did. As you can see the `get_image` task is running successfully, with a green check. But afterward the 500 error pops out.
As a side note, my MCP can run get_code correctly



 

 


Already did. As you can see the `get_image` task is running successfully, with a green check. But afterward the 500 error pops out.
As a side note, my MCP can run get_code correctly



 

 

The same error occurs.

`get_image` works well but i see the 500 error.

The mcp worked well few hours ago but it keeps failing now.

 

Hey ​@Bors ​@fin.chen 

 

Thanks for flagging. Are you still experiencing this issue? This might be related to the Claude API on their end, but if you're continuing to encounter this error, it would be really helpful if you could submit a formal bug report with detailed reproduction steps.

You can do this here: https://help.figma.com/hc/en-us/articles/360041468234-Submit-a-bug-report. The more information you can provide about when and how this occurs, the better our team can investigate.

Thanks for your patience

 

Toku


Hi ​@y_toku 

I changed to gemini 2.5 pro / GTP-4.1 / o4-mini and it did not show 500 error. 
However, now the `get_code` seems to be running infinitely when using gemini 2.5 pro / GPT-4.1 / o4-mini. I’ve filed a bug report. Thanks for the help.


can confirm getting the same bug with the get image running infinitely


Same issue.

Claude Sonnet 4 - 500 error

Gemini 2.5 Pro - 503 error

Claude Sonnet 3.7 - works

GPT 4.1 - infinite loop


Claude Sonnet 3.5 - works

Claude Sonnet 3.7 - works

Claude Sonnet 4 - not working (Sorry, your request failed. Please try again. Request id: 0bde1ecf-2a4b-4669-958b-38e1cc92220f Reason: Server error: 500)

Gemini 2.5 Pro (Preview) - only `get_image` works. the others has stopped after the `get_code` trigger

Gemini 2.5 Flash Preview - not working ( Sorry, your request failed. Please try again. Request id: b4d385a8-3c94-4065-aa8b-f500daaa99b4 Reason: Request Failed: 400 4{ "error": { "code": 400, "message": "Invalid content part type: image_url", "status": "INVALID_ARGUMENT" } } ])

GPT-4.1 - works

GPT-4o - works

o4-mini (Preview)  - works

 

 

I use VS Code and GitHub Copilot Pro Plan.

 
 
 
 

Having the same issue, ‘mcp error’ while trying to ‘get_image’ on Cursor, all models, ‘get_code’ works perfect tho


After some chatting with support, the issue comes from selecting large areas. Which makes the process unable to finish.

 

Upon selecting smaller area, all the models responded normally.

 

As of now, the best usage would be selecting a smaller area to build a small component. And then iterate from that.