second-opinion-mcp-server

An MCP server that queries multiple AI models in parallel via OpenRouter, giving your AI agent a "second opinion" on any question.

Quick Start

1. Get an API key from OpenRouter
2. Add to your MCP client config:

{
  "mcpServers": {
    "second-opinion": {
      "command": "npx",
      "args": ["-y", "second-opinion-mcp-server"],
      "env": {
        "OPENROUTER_API_KEY": "sk-or-..."
      }
    }
  }
}

To run from a local clone instead, build first then point to the dist entry:

{
  "mcpServers": {
    "second-opinion": {
      "command": "node",
      "args": ["/path/to/second_opinion/dist/index.js"],
      "env": {
        "OPENROUTER_API_KEY": "sk-or-..."
      }
    }
  }
}

Tools

get_second_opinion

Sends a query to multiple AI models in parallel and returns all responses as formatted Markdown.

Input:

• query (string, required) — The question or prompt to send to all models
• context (string, optional) — Additional context prepended to the query
• system_prompt (string, optional) — System prompt sent to all models
• models (string[], optional) — Override configured models for this request
• max_tokens (number, optional) — Max response tokens per model
• temperature (number, optional) — Sampling temperature (0-2)

review_code

Send code to multiple AI models for review with a specific focus area.

Input:

• code (string, required) — The code to review
• language (string, optional) — Programming language
• focus (enum, required) — Review focus: security, performance, style, or bugs
• models (string[], optional) — Override configured models
• max_tokens (number, optional) — Max response tokens per model

compare_approaches

Compare two or more technical approaches by sending them to multiple AI models.

Input:

• question (string, required) — The decision or problem to evaluate
• approaches (string[], required, min 2) — The approaches to compare
• models (string[], optional) — Override configured models

fact_check

Fact-check a claim across multiple AI models, prioritizing search-augmented models.

Input:

• claim (string, required) — The claim to fact-check
• models (string[], optional) — Override configured models

follow_up

Drill deeper into a previous second-opinion response. Send the markdown from a prior tool call as context, then ask a follow-up question.

Input:

• previous_responses (string, required) — The markdown output from a prior second-opinion tool call
• question (string, required) — The follow-up question to ask
• system_prompt (string, optional) — System prompt sent to all models
• models (string[], optional) — Override configured models
• max_tokens (number, optional) — Max response tokens per model
• temperature (number, optional) — Sampling temperature (0-2)

list_models

Return configured models with pricing, context length, and capabilities.

Input: none

Output: JSON array where each entry includes: id, name, context_length, max_completion_tokens, modality, input_cost_per_token, output_cost_per_token.

Resources

second-opinion://config

Returns server configuration as JSON (models, timeout, default temperature/max_tokens).

Environment Variables

Default Models

• openai/gpt-5.2
• google/gemini-3-pro-preview
• x-ai/grok-4.1-fast
• perplexity/sonar-reasoning-pro

Override with SECOND_OPINION_MODELS:

"env": {
  "OPENROUTER_API_KEY": "sk-or-...",
  "SECOND_OPINION_MODELS": "openai/gpt-4o,meta-llama/llama-3-70b-instruct"
}

Development

npm install
npm run build
npm test
OPENROUTER_API_KEY=sk-or-... node dist/index.js

License

MIT