Sage MCP Server

A Model Context Protocol (MCP) server for interacting with the Sage Grande Testbed, cyberinfrastructure for AI@Edge. This server provides tools, resources, and prompts for querying sensor data, submitting jobs, and managing Sage nodes.

Quick Setup with Cursor IDE

The easiest way to use Sage MCP is through Cursor IDE with our hosted server:

1. Configure Cursor MCP

Add this to your Cursor MCP configuration file (~/.cursor/mcp.json):

{
  "mcpServers": {
    "sage": {
      "url": "https://mcp.sagecontinuum.org/mcp",
      "headers": {
        "Authorization": "Bearer {username}:{auth_token}"
      }
    }
  }
}

2. Get Your Sage Credentials

1. Visit Sage Portal Access Credentials
2. Sign in with your Sage account
3. Copy your access token
4. Replace {username} with your Sage username
5. Replace {auth_token} with your access token

3. Start Using Sage MCP

Once configured, you can ask Cursor natural language questions about Sage data:

"Show me temperature readings from node W023 in the last hour"
"Find nodes in Chicago with recent camera images"
"What's the highest temperature recorded today across all nodes?"

See the Examples Guide for more query examples.

Features

• Sensor Data Queries: Access environmental, IIO, and other sensor data from Sage nodes
• Job Management: Submit, monitor, and manage edge computing jobs
• Plugin System: Work with Sage plugins and edge applications
• Documentation: Built-in Sage documentation and FAQ system
• Authentication Support: Support for protected data access with user tokens
• Natural Language Queries: Query data using natural language descriptions

Documentation

Comprehensive documentation is available in the docs/ folder:

• Getting Started Guide - Complete setup and usage guide
• Authentication Guide - Detailed authentication implementation
• Custom Functions Guide - How to add custom functionality
• Docker Deployment - Containerized deployment instructions
• LLM Integration - Language model and AI integration details

Authentication

The server supports authentication for accessing protected Sage data through HTTP headers and query parameters only.

Token Format

For protected data access, use the format: username:access_token

• Get your access token from: https://portal.sagecontinuum.org/account/access
• Format: your_username:your_access_token

1. Authorization Header (Recommended)

Use HTTP Authorization header with Basic auth:

# Basic Auth with username:token
curl -H "Authorization: Basic $(echo -n 'username:token' | base64)" \
     "http://localhost:8000/mcp/..."

# Or Bearer token
curl -H "Authorization: Bearer username:token" \
     "http://localhost:8000/mcp/..."

2. Custom Header

Use the custom X-SAGE-Token header:

curl -H "X-SAGE-Token: username:token" \
     "http://localhost:8000/mcp/..."

3. Query Parameter

Pass token as a query parameter:

curl "http://localhost:8000/mcp/...?token=username:token"

Getting Your Access Token

1. Visit Sage Portal Access Credentials
2. Sign in with your Sage account
3. Copy your access token from the credentials section
4. For protected data, ensure you have:
   • A valid Sage account
   • Signed the Data Use Agreement
   • Appropriate permissions for the data you're accessing

Token Requirements

For protected data access, ensure you have:

• Just the token: your_token_here
• Username and token: username:your_token_here

Image Proxy

The server includes an image proxy endpoint that allows authenticated access to Sage images:

HTTP Endpoint

GET /proxy/image?url=<encoded_sage_url>&token=<optional_token>

Authentication Methods (in priority order)

1. Environment Variables (recommended - like the Sage Python examples):

   export SAGE_USER=your_username
   export SAGE_PASS=your_password
   

2. Token parameter in username:password format:

   curl "http://localhost:8000/proxy/image?url=...&token=username:password"
   

3. Bearer token for simple access tokens:

   curl "http://localhost:8000/proxy/image?url=...&token=your_access_token"
   

MCP Tool

Use the get_image_proxy_url() tool to generate proxy URLs:

# Get a proxy URL for a Sage image
proxy_url = get_image_proxy_url("https://storage.sagecontinuum.org/api/v1/data/sage/...")

Features

• Multiple Auth Methods: Environment variables, username:password tokens, or Bearer tokens
• Security: Only allows Sage storage URLs
• Caching: Images are cached for 1 hour for better performance
• Error Handling: Proper HTTP status codes for authentication and access errors

Installation

1. Clone this repository
2. Install dependencies:

pip install -r requirements.txt

# Test that all dependencies are properly installed
python test_dependencies.py

Usage

Starting the Server

python sage_mcp.py

The server will start on http://localhost:8000/mcp by default.

Environment Variables

• MCP_HOST: Host to bind to (default: 0.0.0.0)
• MCP_PORT: Port to bind to (default: 8000)

Example Usage with Authentication

# Method 1: Using MCP tools (recommended)
# Start the server
python sage_mcp.py
# In your MCP client, call: set_authentication_token("username:your_access_token_here")
# Then use any data querying tool

# Method 2: Environment variable
export SAGE_USER_TOKEN="your_access_token_here"
python sage_mcp.py
# All tools will automatically use the token for protected data

Available Tools

Authentication Tools

• set_authentication_token(username, token) - Set your Sage authentication credentials (per-session)
• set_authentication_token_legacy(token) - Set token in legacy format (deprecated)
• get_authentication_status() - Check if an authentication token is set for this session
• list_active_sessions() - List all active authenticated sessions (admin/debug tool)

Sensor Data Tools

• get_node_all_data(node_id, time_range) - Get all sensor data for a node
• get_node_iio_data(node_id, time_range) - Get IIO sensor data
• get_environmental_summary(node_id, time_range) - Get environmental data summary
• get_node_temperature(node_id, sensor_type) - Get temperature data
• get_temperature_summary(time_range, sensor_type) - Get temperature summary
• search_measurements(pattern, node_id, time_range) - Search for measurements

Node Information Tools

• list_available_nodes(time_range) - List active Sage nodes
• get_node_info(node_id) - Get detailed node information
• list_all_nodes() - List all Sage nodes
• get_sensor_details(sensor_type) - Get sensor specifications

Job Management Tools

• submit_sage_job(job_name, nodes, plugin_image, ...) - Submit custom jobs
• submit_plugin_job(plugin_type, job_name, nodes) - Submit pre-configured plugin jobs
• check_job_status(job_id) - Check job status
• query_job_data(job_name, node_id, time_range) - Query job output data

Geographic Tools

• get_nodes_by_location(location) - Find nodes by geographic location
• get_measurement_stat_by_location(location, measurement_type, stat, ...) - Get statistics by location

Plugin Tools

• find_plugins_for_task(task_description) - Find plugins for a task
• get_plugin_data(plugin_id, nodes, time_range) - Query plugin data
• query_plugin_data_nl(query) - Natural language plugin queries

Image and Cloud Data Tools

• get_cloud_images(time_range, node_id) - Get cloud/sky images
• get_image_data(time_range, node_id, plugin_pattern) - Get image data
• get_image_proxy_url(sage_url) - Get a proxy URL for accessing Sage images with authentication

Documentation Tools

• ask_sage_docs(question) - Ask questions about Sage documentation
• sage_faq(topic) - Get FAQ answers
• search_sage_docs(query) - Search documentation

Resources

• query://{plugin} - Query data for specific plugins
• stats://temperature - Temperature statistics across nodes

Prompts

• getting_started_guide() - Interactive guide for new users
• plugin_development_guide() - Guide for creating plugins
• data_analysis_guide() - Guide for data analysis
• troubleshooting_guide() - Troubleshooting help

Docker Deployment

See DOCKER_DEPLOY.md for containerized deployment instructions.

Development

The server is built using:

• FastMCP - MCP server framework
• sage-data-client - Sage data access
• pandas - Data processing

Extending with Custom Functions

Want to add your own custom MCP tools to the Sage server? You can easily fork this repository and add custom endpoints:

@mcp.tool()
def my_custom_analysis(data_query: str, analysis_type: str = "basic") -> str:
    """Perform custom analysis on Sage data"""
    # Your custom logic here
    # Access Sage data using the existing data_service
    # Return formatted results
    return "Analysis results..."

For detailed instructions on forking, adding custom functions, and deploying your enhanced server, see our Custom Functions Guide.

Testing

Testing scripts are located in the tests/ folder:

# Test authentication functionality
python tests/test_auth.py

# Test dependencies
python tests/test_dependencies.py

# Test image proxy functionality
python tests/test_image_proxy.py

# Test Sage authentication
python tests/test_sage_auth.py

# Run server tests
python tests/test_server.py

The authentication test will verify all supported authentication methods (headers and query parameters) work correctly.

License

This project is licensed under the MIT License.