Creating Agent Cards
Learn how to create Agent Cards that define your AI agent’s capabilities and enable agent-to-agent collaboration.
What You’ll Build
In this guide, you’ll create a Weather Assistant Agent Card with:
- Basic agent information
- Skills definitions
- Capabilities configuration
- MCP server integration
Time Required: 10 minutes
Step 1: Navigate to Agent Cards
- Login to MA²D
- Click “Agent Cards” in the sidebar
- Click ”+ New Agent Card”
Step 2: Basic Information
Fill in the agent details:
Name:
weather-assistantVersion:
1.0.0Description:
Intelligent weather assistant providing real-time weather data, forecasts, and weather-related insights for cities worldwide. Supports multiple temperature units and natural language queries.Author:
Your Organization NameLicense:
Apache-2.0Click “Next” or scroll down.
Step 3: Define Skills
Skills describe what your agent can do. Click ”+ Add Skill”.
Skill 1: Get Weather
Name:
get_weatherDescription:
Get current weather conditions including temperature, humidity, wind speed, and weather conditions for any city worldwideParameters:
{
"city": "string - City name",
"units": "string - celsius or fahrenheit (optional)"
}Returns:
Weather object with current conditionsMCP Server:
weather-api(Select from your servers)
MCP Tool:
get_weatherClick “Save Skill”.
Skill 2: Get Forecast (Optional)
Add another skill:
Name: get_forecast
Description: “Get 7-day weather forecast with daily conditions”
Parameters: {"city": "string", "days": "number (1-7)"}
Returns: “Array of daily forecast objects”
MCP Server: weather-api
MCP Tool: get_forecast
Step 4: Configure Capabilities
Define what your agent supports:
Languages:
- English
- Spanish
- French
Domains:
weather, meteorology, climateInput Modes:
- Text
- Voice
- Image
Output Modes:
- Text
- JSON
- Speech
Technical Capabilities:
- Streaming
- Async
- Rate Limit:
100 requests/minute
Step 5: Set Invocation Details
Configure how agents call your agent:
Protocol:
MCPEndpoint:
https://ma2d.vercel.app/api/platform/YOUR_SERVER_ID/mcp(Auto-filled from linked MCP server)
Authentication:
- Type:
Bearer Token(optional) - Required:
No(for public agents)
Method:
POSTFormat:
json-rpcStep 6: Add Examples
Provide sample interactions:
Example 1
Skill:
get_weatherDescription:
Get weather for San FranciscoInput:
{
"city": "San Francisco",
"units": "fahrenheit"
}Output:
{
"city": "San Francisco",
"temperature": 65,
"units": "fahrenheit",
"conditions": "Partly cloudy",
"humidity": 72,
"wind_speed": 12
}Click “Add Example” and save.
Add 2-3 more examples for different scenarios.
Step 7: Review and Save
- Review all information
- Click “Create Agent Card”
- Your Agent Card is now created!
Step 8: Test Integration
Verify the Agent Card links to your MCP server:
- Click on your Agent Card
- Go to “Test” tab
- Select a skill
- Click “Test Skill”
- Verify response from MCP server
Step 9: Publish to Exchange
When ready, publish your Agent Card:
- Ensure linked MCP server is published first
- Click “Publish to Exchange”
- Configure Exchange metadata
- Click “Publish”
Your Agent Card is now discoverable by other agents!
Best Practices
- Clear Descriptions - Be specific about capabilities
- Complete Skills - Document all parameters
- Realistic Examples - Use real-world scenarios
- Link to MCP - Connect to published MCP servers
- Keep Updated - Update when capabilities change
Next Steps
- Agent Cards Concept - Deep dive into A2A
- Agent Card Spec - Complete schema
- Publishing Guide - Share your work
Congratulations! You’ve created your first Agent Card! 🎉