MCP Client and Server Architecture
What is the Model Context Protocol?
The Model Context Protocol (MCP) enables AI applications to access external tools, data sources, and prompts. It serves as a bridge between AI models and the outside world, allowing language models to interact with application-specific resources and functionality.
MCP provides several core capabilities:
- Tools: Functions that can be invoked by LLM applications
- Resources: Data that can be accessed by LLM applications
- Prompts: Templates for LLM interactions
- Notifications: Event-based communication between server and client
In short, MCP gives LLM applications a standardized way to interact with external resources and tools.
Architecture
The AJ MCP SDK is structured around several core components that work together to implement the MCP specification:
The architecture follows a client-server model with flexible transport options. The client exchanges JSON-RPC messages over standard I/O pipes or the legacy HTTP/SSE transport implemented by this version.
Stdio Mode vs SSE Mode Comparison
| Feature | Stdio Mode | SSE Mode |
|---|---|---|
| Deployment | Local subprocess | Independent server |
| Use Case | Local development | Distributed environments |
| Configuration | Subprocess command required | HTTP endpoints required |
| Multi-Client Support | Not supported | Supported |
| Network Requirement | None | Requires network connection |