Skip to main content
The Universal MCP system is a framework designed to enable AI models to interact with external systems and APIs. It achieves this through a core set of components:
  • Applications
  • Actions
  • Integration
  • Triggers

​
Applications

Applications form the foundation for connecting to external services. They manage aspects like:
  • Authentication
  • API communication
  • Defining available operations
  • Handling responses and errors
The system provides base classes to simplify the creation of new applications:
  • BaseApplication (abstract base class)
  • APIApplication (for HTTP APIs)
  • GraphQLApplication (for GraphQL APIs)
Tip: Creating a new application involves setting up a specific package structure and implementing a class that inherits from one of these base classes.

​
Actions

Actions define the specific operations that an application can perform. These allow AI models to:
  • Execute external functions
  • Call APIs
  • Process data
  • Interact with real-world systems
Key features of Actions:
  • Dynamic function calling
  • Structured input/output
  • Integration with various authentication methods
  • Error handling
  • Tool management
How to create custom actions:
  • Define methods within an application class
  • Include type hints and docstrings
  • Implement error handling
  • Register them in the list_tools() method

​
Integration

Integration is the mechanism that allows Applications and Actions to work together. It handles:
  • Authentication
  • Credential management
  • API communication
  • Error handling
The system supports several integration types, including:
  • API Key
  • OAuth
  • Basic Auth

​
Triggers

Triggers act as a notification system, enabling AI agents to respond dynamically to external events from integrated applications (like GitHub or Gmail).
When specific events occur in these applications, triggers notify the agents, allowing for real-time, event-driven workflows.