MCP Setup

MCP Setup

This page shows how to install and run the AutoVio MCP server so MCP clients (like Claude Desktop) can call the AutoVio API.

Installation

From the AutoVio-MCP directory:

npm install
npm run build

This compiles the TypeScript sources into dist/ and prepares the CLI entry point.

Running the server

You can run the MCP server directly with Node:

# Default (env or default baseUrl/token)
node dist/index.js

# With config file
node dist/index.js --config examples/config.example.json

# Development (watch)
npm run dev

The entry point (src/index.ts) loads configuration, sets the log level, and then calls startServer with stdio transport.

Configuration sources and precedence

Configuration is merged from four sources with this priority (highest first):

1. CLI parameters
2. Environment variables
3. Config file (JSON)
4. Defaults

1. CLI parameters

CLI flags cover AutoVio connection, logging, and the 4 AI model+key pairs:

Flag	Description
--config	Path to JSON config file.
--autovio-base-url	AutoVio API base URL.
--autovio-api-token	AutoVio API token.
--vision-model	Vision model (e.g. gemini-2.0-flash-exp).
--vision-api-key	Vision API key.
--llm-model	LLM model (e.g. gemini-2.5-flash).
--llm-api-key	LLM API key.
--image-model	Image model (e.g. gemini-2.5-flash-image).
--image-api-key	Image API key.
--video-model	Video model (e.g. veo-3.0-generate-001).
--video-api-key	Video API key.
--log-level	debug | info | warn | error.
--enable-resources	true | false.
--enable-prompts	true | false.

CamelCase variants like --autovioBaseUrl are also accepted.

2. Environment variables

The server reads the following env variables:

• AUTOVIO_BASE_URL
• AUTOVIO_API_TOKEN
• AUTOVIO_VISION_MODEL, AUTOVIO_VISION_API_KEY
• AUTOVIO_LLM_MODEL, AUTOVIO_LLM_API_KEY
• AUTOVIO_IMAGE_MODEL, AUTOVIO_IMAGE_API_KEY
• AUTOVIO_VIDEO_MODEL, AUTOVIO_VIDEO_API_KEY
• AUTOVIO_LOG_LEVEL
• AUTOVIO_MCP_CONFIG (JSON config string)

3. Config file

You can pass a JSON config via --config or load it through env. Example (examples/config.example.json):

{
  "server": {
    "name": "autovio-mcp-server",
    "version": "1.0.0"
  },
  "autovio": {
    "baseUrl": "http://localhost:3001",
    "apiToken": "YOUR_AUTOVIO_API_TOKEN"
  },
  "providers": {
    "vision": {
      "model": "gemini-2.0-flash-exp",
      "apiKey": "YOUR_VISION_API_KEY"
    },
    "llm": {
      "model": "gemini-2.5-flash",
      "apiKey": "YOUR_LLM_API_KEY"
    },
    "image": {
      "model": "gemini-2.5-flash-image",
      "apiKey": "YOUR_IMAGE_API_KEY"
    },
    "video": {
      "model": "veo-3.0-generate-001",
      "apiKey": "YOUR_VIDEO_API_KEY"
    }
  },
  "features": {
    "enableResources": true,
    "enablePrompts": false,
    "cacheResponses": false,
    "logLevel": "info"
  }
}

4. Default structure

At runtime the merged config has this shape:

Section	Fields	Description
server	name, version	MCP server metadata.
autovio	baseUrl, apiToken	AutoVio REST API connection.
providers	vision, llm, image, video	Model+API key pairs for each AI category.
features	enableResources, enablePrompts, cacheResponses, logLevel	MCP feature flags and logging.

Each provider entry is:

Field	Type	Description
model	string	Model name (e.g. gemini-2.0-flash-exp).
apiKey	string	API key for that model.

The server derives the provider ID from the model name (e.g. gemini, claude, openai) before calling the AutoVio API.

Requirements

• Node.js >= 18 (enforced by engines.node in package.json)
• A running AutoVio backend (see Installation)
• An AutoVio API token or user credentials
• API keys for the AI providers you want to use

See also

• MCP Overview
• Claude Desktop Integration
• MCP Tools Reference