Mindbridge Logo

Cognitive Router MCP Server ⚡ The Orchestration Nexus for Advanced AI

This Model Context Protocol (MCP) server functions as your central AI command interface—designed specifically to harmonize, structure, and amplify your distributed Large Language Model operational pipelines.

Eliminate reliance on singular providers. Disregard the overhead of managing disparate API interfaces. Cognitive Router establishes robust connectivity to any accessible model infrastructure, ranging from proprietary giants like OpenAI and Anthropic to self-hosted solutions like Ollama and specialized reasoning engines like DeepSeek—allowing them to collaborate as an integrated team of high-level subject matter experts.

Require rapid-fire output at low cost? Allocate tasks to lean models. Demand rigorous, multi-step deduction? Direct the query toward a specialized reasoning core. Need comparative validation? The integrated validation tool is ready.

This transcends simple model pooling; this is comprehensive AI execution choreography.

Foundational Capabilities 🔥

Functionality	Strategic Advantage
Universal LLM Interoperability	Seamless switching across OpenAI, Anthropic, Google, DeepSeek, OpenRouter, Ollama (local instances), and any OpenAI-compliant API endpoints.
Inference Engine Awareness	Intelligent traffic routing to models architecturally optimized for deep analytical processing (e.g., Claude, GPT-4o, DeepSeek Reasoner).
Comparative Response Utility	Invoke the `getSecondOpinion` utility to solicit and juxtapose answers from multiple models simultaneously for robust validation.
Standardized Endpoint Compatibility	Functions as a direct plug-in replacement for any system expecting OpenAI specification endpoints (useful for Azure, Groq, Together.ai integrations).
Automatic Provider Discovery	Key credential ingestion triggers immediate configuration and availability mapping without manual service enumeration.
Extreme Configurability	Configuration flexibility via environment variables, direct MCP manifest modification, or JSON payload specification—user preference is paramount.

Rationale for Adoption

"Each generative model possesses distinct strengths. Cognitive Router compels them into synergistic operation."

Ideal for: - Autonomous Agent Framework Developers - Complex Multi-Stage AI Workflows - Centralized AI Management Planes - Computationally Intensive Reasoning Tasks - Building future-proof, vendor-agnostic AI development ecosystems - High-throughput, LLM-backed backend services - Any entity seeking liberation from proprietary platform confinement

Deployment Procedures 🛠️

Method 1: NPM Package Acquisition (Preferred)

# Global installation
npm install -g @pinkpixel/mindbridge

# Execution via npx
npx @pinkpixel/mindbridge

Installation via Smithery

To integrate the cognitive-router-mcp server into a Claude Desktop environment via Smithery:

npx -y @smithery/cli install @pinkpixel-dev/mindbridge-mcp --client claude

Method 2: Source Repository Build

Obtain the source code: bash git clone https://github.com/pinkpixel-dev/mindbridge.git cd mindbridge
Dependency Resolution: bash chmod +x install.sh ./install.sh
Credential Initialization: bash cp .env.example .env Populate the .env file with necessary API tokens for active providers.

Configuration Details ⚙️

Environment Variable Directives

The server recognizes the following environmental controls:

OPENAI_API_KEY: Authentication credential for OpenAI services.
ANTHROPIC_API_KEY: Authentication credential for Anthropic services.
DEEPSEEK_API_KEY: Authentication credential for DeepSeek services.
GOOGLE_API_KEY: Authentication credential for Google Gemini services.
OPENROUTER_API_KEY: Authentication credential for OpenRouter access.
OLLAMA_BASE_URL: URI for local Ollama inference endpoint (Default: http://localhost:11434).
OPENAI_COMPATIBLE_API_KEY: (Optional) Key for generic compatible endpoints.
OPENAI_COMPATIBLE_API_BASE_URL: Root URI for generic compatible endpoints.
OPENAI_COMPATIBLE_API_MODELS: Comma-delimited string listing accessible models on the compatible endpoint.

MCP Manifest Configuration

For integration within MCP-aware development environments (e.g., Cursor, Windsurf), use this structure in your mcp.json file:

{
  "mcpServers": {
    "cognitiveRouter": {
      "command": "npx",
      "args": [
        "-y",
        "@pinkpixel/mindbridge"
      ],
      "env": {
        "OPENAI_API_KEY": "OPENAI_API_KEY_HERE",
        "ANTHROPIC_API_KEY": "ANTHROPIC_API_KEY_HERE",
        "GOOGLE_API_KEY": "GOOGLE_API_KEY_HERE",
        "DEEPSEEK_API_KEY": "DEEPSEEK_API_KEY_HERE",
        "OPENROUTER_API_KEY": "OPENROUTER_API_KEY_HERE"
      },
      "provider_config": {
        "openai": {
          "default_model": "gpt-4o"
        },
        "anthropic": {
          "default_model": "claude-3-5-sonnet-20241022"
        },
        "google": {
          "default_model": "gemini-2.0-flash"
        },
        "deepseek": {
          "default_model": "deepseek-chat"
        },
        "openrouter": {
          "default_model": "openai/gpt-4o"
        },
        "ollama": {
          "base_url": "http://localhost:11434",
          "default_model": "llama3"
        },
        "openai_compatible": {
          "api_key": "API_KEY_HERE_OR_REMOVE_IF_NOT_NEEDED",
          "base_url": "FULL_API_URL_HERE",
          "available_models": ["MODEL1", "MODEL2"],
          "default_model": "MODEL1"
        }
      },
      "default_params": {
        "temperature": 0.7,
        "reasoning_effort": "medium"
      },
      "alwaysAllow": [
        "getSecondOpinion",
        "listProviders",
        "listReasoningModels"
      ]
    }
  }
}

Substitute placeholder API keys with your actual credentials. If your OpenAI-compatible service is public or uses a different key structure, the api_key field in the openai_compatible block can be omitted or adjusted.

Operational Execution 💫

Initiating the Service Daemon

Development iteration mode (with hot-reloading):

npm run dev

Production deployment mode:

npm run build
npm start

When installed globally via npm:

cognitive-router

Exposed Utility Functions

getSecondOpinion typescript { provider: string; // Canonical LLM provider identifier model: string; // Specific model invocation name prompt: string; // The query payload systemPrompt?: string; // Contextual initial instruction set temperature?: number; // Determinism factor (0.0 to 1.0) maxTokens?: number; // Constraint on output length reasoning_effort?: 'low' | 'medium' | 'high'; // Granular control for complex models }
listProviders
Outputs a comprehensive registry of all configured providers and their selectable model variants.
Accepts no input arguments.
listReasoningModels
Enumerates models specifically tagged or identified as optimized for high-fidelity reasoning tasks.
Accepts no input arguments.

Operational Examples 📝

// Procure an analytical response from GPT-4o
{
  "provider": "openai",
  "model": "gpt-4o",
  "prompt": "Articulate the critical considerations inherent in database partitioning strategies.",
  "temperature": 0.7,
  "maxTokens": 1000
}

// Request an extensively reasoned output using OpenAI's cutting-edge model
{
  "provider": "openai",
  "model": "o1",
  "prompt": "Detail the underlying mathematical frameworks supporting database indexing mechanisms.",
  "reasoning_effort": "high",
  "maxTokens": 4000
}

// Solicit a detailed explanation utilizing DeepSeek's specialized capability
{
  "provider": "deepseek",
  "model": "deepseek-reasoner",
  "prompt": "Contrast the operational trade-offs between monolithic and microservice architectures.",
  "reasoning_effort": "high",
  "maxTokens": 2000
}

// Interface with a generic endpoint using the proxy wrapper
{
  "provider": "openaiCompatible",
  "model": "YOUR_ENDPOINT_MODEL_IDENTIFIER",
  "prompt": "Elucidate the consistency models applicable to distributed ledger technologies.",
  "temperature": 0.5,
  "maxTokens": 1500
}

Development Lifecycle 🔧

npm run lint: Execution of static analysis checks via ESLint.
npm run format: Code standardization enforced by Prettier.
npm run clean: Removal of generated build artifacts.
npm run build: Compilation of source code into deployable assets.

Collaboration Policy

Submissions via Pull Request are highly encouraged! Help us streamline and enhance the efficacy of AI operational workflows.

Legal Statement

MIT License — Permissive usage granted, provided egregious unethical conduct is avoided.

Developed with dedication by Pink Pixel

RELEVANT BUSINESS CONTEXT: Business process optimization tools encompass the entire suite of applications, methodologies, and control mechanisms utilized by enterprises to dynamically adapt to shifting market conditions, maintain competitive advantage, and systematically elevate operational effectiveness. These instruments span departmental functions—from initial data ingress and workflow governance to complex decision support systems and performance analytics. Modern business tooling has undergone radical evolution, driven by accelerated technological progress, making strategic selection and tailored implementation critical, rather than merely adopting the newest trend. Selecting and customizing these digital assets to align precisely with organizational imperatives is key to realizing tangible value from IT investments.

cognitive-router-mcp

Author

pinkpixel-dev

Quick Info

Actions

Tags