What happened?

Summary

gemini fails for every conversation prompt with a 400 error when the Notion MCP server is enabled. As soon as I disable that MCP entry in settings.json, the CLI works normally again.

Observed error

╭────────╮
│ > hi │
╰────────╯

✕ [API Error: {"error":{"message":"{\n "error": {\n "code": 400,\n "message": "A FunctionDeclaration in the request exceeds the maximum allowed
nesting depth.",\n "status": "INVALID_ARGUMENT"\n }\n}\n","code":400,"status":"Bad Request"}}]

╭───────────╮
│ > hello │
╰───────────╯

✕ [API Error: {"error":{"message":"{\n "error": {\n "code": 400,\n "message": "A FunctionDeclaration in the request exceeds the maximum allowed
nesting depth.",\n "status": "INVALID_ARGUMENT"\n }\n}\n","code":400,"status":"Bad Request"}}]

Minimal settings excerpt (before)

{
  "mcpServers": {
    "notionMCP": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.notion.com/mcp"]
    }
  }
}

Workaround (after)

Commenting out or removing just the Notion MCP entry makes the CLI work again:

{
  "mcpServers": {
    // "notionMCP": {
    //   "command": "npx",
    //   "args": ["-y", "mcp-remote", "https://mcp.notion.com/mcp"]
    // }
  }
}

What did you expect to happen?

Prompts should succeed even with Notion MCP enabled.

Client information

• CLI Version: 0.1.18-nightly.250810.c632ec8b
• Git Commit: c632ec8
• Operating System: linux v20.19.3
• Sandbox Environment: no sandbox
• Model Version: gemini-2.5-pro
• Memory Usage: 138.0 MB

Login information

API key

Anything else we need to know?

• The failure reproduces consistently with any simple prompt.

• Disabling only the Notion MCP (leaving other MCP servers intact) removes the error, suggesting the failure is triggered by this server’s FunctionDeclarations (likely exceeding the API’s nesting-depth constraints).