Explorium MCP icon

Explorium MCP

Make requests to OpenAI with Explorium MCP integration

Overview

This node integrates OpenAI's language models with Explorium's MCP (Managed Content Platform) tool. It sends a user-defined prompt to an OpenAI model while leveraging Explorium's MCP for enhanced processing or moderation. The node supports requiring approval for using the MCP tool, which can be configured per execution.

Common scenarios include:

  • Generating AI responses that must pass through Explorium's content moderation or enrichment layer.
  • Using Explorium MCP to control or audit AI-generated content before final output.
  • Automating workflows where AI outputs require conditional approval based on organizational policies.

Practical example:

  • A marketing team uses this node to generate product descriptions via OpenAI but requires automatic approval checks through Explorium MCP to ensure compliance with brand guidelines.

Properties

Name Meaning
Model The OpenAI model to use. Example: "gpt-4.1".
Prompt The text prompt sent to the AI model for generating a response.
Require Approval Whether to require approval for MCP tool usage. Options: "Never", "Always", "Auto".

Output

The node outputs JSON data containing the AI assistant's response text under the response field. The structure is:

{
  "response": "string"
}
  • If the API returns an error, the output will contain an error field with the error message.
  • The node does not output binary data.

Dependencies

  • Requires valid credentials for both OpenAI API and Explorium API.
  • Needs network access to:
    • OpenAI endpoint: https://api.openai.com/v1/responses
    • Explorium MCP server: https://mcp.explorium.ai/sse
  • The node uses these credentials to authenticate requests to both services.

Troubleshooting

  • Common issues:
    • Invalid or missing API keys for either OpenAI or Explorium APIs will cause authentication failures.
    • Network connectivity problems to the specified endpoints.
    • Improper prompt formatting may lead to unexpected or empty responses.
  • Error messages:
    • Errors returned from the OpenAI API or Explorium MCP are captured and shown in the output under the error field.
    • To resolve errors, verify API keys, check network connectivity, and ensure the prompt is correctly set.
  • The node supports continuing on failure if enabled, allowing partial success in batch executions.

Links and References

Discussion