mcp-agent8

MCP Server for Agent8

A server implementing the Model Context Protocol (MCP) to support Agent8 SDK development. Developed with TypeScript and pnpm, supporting stdio, SSE, and streamable-http transports.

Features

This Agent8 MCP Server implements the following MCP specification capabilities:

Prompts

• System Prompt for Agent8 SDK: Provides optimized guidelines for Agent8 SDK development through the system-prompt-for-agent8-sdk prompt template.

Tools

• Code Examples Search: Retrieves relevant Agent8 game development code examples from a vector database using the search_code_examples tool.
• Game Resource Search: Searches for game development assets (sprites, animations, sounds, etc.) using semantic similarity matching via the search_game_resources tool.
• Asset Generation: Comprehensive toolset for game asset creation:
  • Images: Generate 2D game assets using the image_asset_generate tool
  • Cinematics: Create cinematic sequences with cinematic_asset_generate tool
  • Audio: Generate music tracks and sound effects with music_generate and sfx_generate tools
  • Skyboxes: Create 360° environmental backgrounds with skybox_generate tool
  • Support Tools: Status checking, result retrieval, and wait utilities for asynchronous generation

Installation

# Install dependencies
pnpm install

# Build
pnpm build

Using Docker

You can run this application using Docker in several ways:

Option 1: Pull from GitHub Container Registry (Recommended)

# Pull the latest image
docker pull ghcr.io/planetarium/mcp-agent8:latest

# Run the container
docker run -p 3333:3333 --env-file .env ghcr.io/planetarium/mcp-agent8:latest

Option 2: Build Locally

# Build the Docker image
docker build -t agent8-mcp-server .

# Run the container with environment variables
docker run -p 3333:3333 --env-file .env agent8-mcp-server

Docker Environment Configuration

There are three ways to configure environment variables when running with Docker:

1. Using --env-file (Recommended):

   # Create and configure your .env file first
   cp .env.example .env
   nano .env
   
   # Run with .env file
   docker run -p 3000:3000 --env-file .env agent8-mcp-server
   

2. Using individual -e flags:

   docker run -p 3000:3000 \
     -e SUPABASE_URL=your_supabase_url \
     -e SUPABASE_SERVICE_ROLE_KEY=your_service_role_key \
     -e OPENAI_API_KEY=your_openai_api_key \
     -e MCP_TRANSPORT=sse \
     -e PORT=3000 \
     -e LOG_LEVEL=info \
     agent8-mcp-server
   

3. Using Docker Compose (for development/production setup):

   The project includes a pre-configured docker-compose.yml file with:

   • Automatic port mapping from .env configuration
   • Environment variables loading
   • Volume mounting for data persistence
   • Container auto-restart policy
   • Health check configuration

   To run the server:

   docker compose up
   

   To run in detached mode:

   docker compose up -d
   

Required Environment Variables:

• SUPABASE_URL: Supabase URL for database connection
• SUPABASE_SERVICE_ROLE_KEY: Supabase service role key for authentication
• OPENAI_API_KEY: OpenAI API key for AI functionality

The Dockerfile uses a multi-stage build process to create a minimal production image:

• Uses Node.js 20 Alpine as the base image for smaller size
• Separates build and runtime dependencies
• Only includes necessary files in the final image
• Exposes port 3000 by default

Usage

Command Line Options

# View help
pnpm start --help

# View version information
pnpm start --version

Supported options:

• --debug: Enable debug mode
• --transport <type>: Transport type (stdio, sse, or streamable-http), default: stdio
• --port <number>: Port to use for SSE or HTTP transport, default: 3000
• --log-destination <dest>: Log destination (stdout, stderr, file, none)
• --log-file <path>: Path to log file (when log-destination is file)
• --log-level <level>: Log level (debug, info, warn, error), default: info
• --env-file <path>: Path to .env file

Using Environment Variables

The server supports configuration via environment variables, which can be set directly or via a .env file.

1. Create a .env file in the project root (see .env.example for reference):

# Copy the example file
cp .env.example .env

# Edit the .env file with your settings
nano .env

2. Run the server (it will automatically load the .env file):

pnpm start

3. Or specify a custom path to the .env file:

pnpm start --env-file=/path/to/custom/.env

Configuration Priority

The server uses the following priority order when determining configuration values:

1. Command line arguments (highest priority)
2. Environment variables (from .env file or system environment)
3. Default values (lowest priority)

This allows you to set baseline configuration in your .env file while overriding specific settings via command line arguments when needed.

Supported Environment Variables

Tool Activation Priority: The tool activation settings follow this priority order:

1. Individual tool settings (e.g., ENABLE_CODE_EXAMPLE_SEARCH_TOOL)
2. Asset type settings (e.g., ENABLE_IMAGE_GENERATION_TOOLS, ENABLE_CINEMATIC_GENERATION_TOOLS)
3. Tool group settings (e.g., ENABLE_VECTOR_SEARCH_TOOLS, ENABLE_ASSET_GENERATE_TOOLS)
4. Global tool setting (ENABLE_ALL_TOOLS)

Individual settings always override group settings, and group settings override the global setting. When individual settings are explicitly set, they take precedence over their parent settings.

Important: To enable only specific tools, you should set all higher-level settings to false and only enable the specific tools you need. This approach provides a more consistent and predictable configuration.

Examples:

# Enable only vector search tools
ENABLE_ALL_TOOLS=false
ENABLE_VECTOR_SEARCH_TOOLS=true

# Enable only image generation tool, disable all others
ENABLE_ALL_TOOLS=false
ENABLE_ASSET_GENERATE_TOOLS=false
ENABLE_IMAGE_GENERATION_TOOLS=true

# Enable only code example search tool, disable all others
ENABLE_ALL_TOOLS=false
ENABLE_VECTOR_SEARCH_TOOLS=false
ENABLE_CODE_EXAMPLE_SEARCH_TOOL=true

# Enable only cinematic and audio generation tools
ENABLE_ALL_TOOLS=false
ENABLE_ASSET_GENERATE_TOOLS=false
ENABLE_CINEMATIC_GENERATION_TOOLS=true
ENABLE_AUDIO_GENERATION_TOOLS=true

Using Stdio Transport

# Build and run
pnpm build
pnpm start --transport=stdio

Using SSE Transport

# Build and run (default port: 3000)
pnpm build
pnpm start --transport=sse --port=3000

Using Streamable HTTP Transport

# Build and run (default port: 3000)
pnpm build
pnpm start --transport=streamable-http --port=3000

Debug Mode

# Run in debug mode
pnpm start --debug

Available Prompts

• systemprompt-agent8-sdk

Client Integration

Using with Claude Desktop

1. Add the following to Claude Desktop configuration file (claude_desktop_config.json):

{
  "mcpServers": {
    "Agent8": {
      "command": "npx",
      "args": ["--yes", "agent8-mcp-server"]
    }
  }
}

1. Restart Claude Desktop

Adding New Prompts

Add new prompts to the registerSamplePrompts method in the src/prompts/provider.ts file.

License

MIT