Summary

The OpenRouter MCP server on mcp.pipedream.net has a schema validation error that prevents sending chat completion and completion requests to OpenRouter API.

Issue Details

Expected Behavior

According to OpenRouter's API documentation, requests should use:

• model (string) - The model ID to use
• messages (array) - For chat completions
• prompt (string) - For completions

Actual Behavior

The MCP server is expecting a models parameter (plural, as an array) which doesn't exist in OpenRouter's API specification. This causes all chat completion and completion requests to fail with a validation error.

Error Message

{
  "error": "AI_InvalidToolArgumentsError",
  "message": "Invalid arguments for tool",
  "validation_errors": [
    {
      "code": "invalid_type",
      "expected": "array",
      "received": "undefined",
      "path": ["models"],
      "message": "Required"
    }
  ]
}

Steps to Reproduce

1. Configure Claude Desktop to use the Pipedream OpenRouter MCP server
2. Use the OPENROUTER-SEND-CHAT-COMPLETION-REQUEST function
3. Provide valid parameters per OpenRouter's API:

   {
     "model": "google/gemini-2.5-pro-preview-05-06",
     "messages": [{"role": "user", "content": "Hello"}]
   }
4. Observe validation error expecting models array parameter

Root Cause Analysis

The MCP server appears to have incorrect schema validation that expects:

• models: array (incorrect - not part of OpenRouter API)

Instead of:

• model: string (correct - per OpenRouter documentation)

Affected Functions

• OPENROUTER-SEND-CHAT-COMPLETION-REQUEST
• OPENROUTER-SEND-COMPLETION-REQUEST

Working Functions

• OPENROUTER-RETRIEVE-AVAILABLE-MODELS (works correctly)

Environment

• Platform: Claude Desktop with MCP support
• MCP Server: https://mcp.pipedream.net
• OpenRouter API Version: v1

Suggested Fix

Update the schema validation for the affected functions to expect:

• model (singular, string) instead of models (plural, array)
• Align with OpenRouter's official API specification: https://openrouter.ai/docs/api-reference

Additional Context

• The model listing function works correctly, confirming the connection to OpenRouter is functional
• This appears to be a simple schema mismatch rather than a deeper integration issue
• Users cannot use any OpenRouter models for actual inference through the MCP server due to this bug