@blue-trianon/mcp-llm-inference

MCP tool for LLM inference — chat completions, text generation, and model discovery via the L402 API.

Features

• Chat completions with multi-turn message history
• Single-prompt text generation
• List available models
• Optional streaming support
• Configurable base URL via environment variable
• Powered by L402 micropayments

MCP Server Configuration

Add to your MCP client configuration:

{
  "mcpServers": {
    "llm-inference": {
      "command": "npx",
      "args": ["-y", "@blue-trianon/mcp-llm-inference"],
      "env": {
        "NAUTDEV_BASE_URL": "https://api.nautdev.com"
      }
    }
  }
}

Or if installed globally:

{
  "mcpServers": {
    "llm-inference": {
      "command": "mcp-llm-inference"
    }
  }
}

Tools

chat

Send a chat completion request to an LLM.

| Parameter | Type | Required | Description | |-----------|---------|----------|------------------------------------------------| | model | string | yes | Model identifier (e.g. llama3, mistral, gemma) | | messages | array | yes | Array of {role, content} message objects | | stream | boolean | no | Stream the response (default: false) |

generate

Generate text from a single prompt.

| Parameter | Type | Required | Description | |-----------|---------|----------|------------------------------------------------| | model | string | yes | Model identifier (e.g. llama3, mistral, gemma) | | prompt | string | yes | Text prompt for generation | | stream | boolean | no | Stream the response (default: false) |

models

List all available LLM models.

No parameters required.

Pricing

Requests are metered via L402 micropayments. See Blue-Trianon-Ventures for pricing details.

License

MIT