Known issues

This document provides details and workarounds for known issues with Google and Google Cloud remote Model Context Protocol (MCP) servers.

Tool issues

The following known issues affect specific Google or Google Cloud MCP tools.

Compute Engine list_images tool returns permission denied

The Compute Engine list_images tool fails with a permission error when querying public images like fedora-cloud.

Client-specific issues

The following issues involve problems with specific MCP clients.

Gemini CLI gives API error

The following issue occurs when the name of the MCP server is longer than 64 characters or contains characters other than alphanumeric characters, underscores, dots, colons, or dashes:

[API Error: {"error":{"message":"{\"error\": {\"code\": 400, \"status\": 400, \"message\": \"* GenerateContentRequest.tools[0].function_declarations[14].name: Invalid
function name. Must start with a letter or an underscore. Must be alphanumeric (a-z, A-Z, 0-9), underscores (_), dots (.), colons (:), or dashes (-), with a maximum length
  of 64.\\n*

To resolve this issue, change the name of the MCP server in your Gemini extension file to comply with the required naming conventions. Gemini CLI extension files are typically located in: ~/.gemini/extensions/EXT_NAME/gemini-extension.json where ~/ is your home directory and EXT_NAME is the name you gave the extension. For more information, see Configure MCP in an AI application

What's next