Crawlab MCP Server

--- description: MCP Server/Client Communication Architecture globs: alwaysApply: false --- ## MCP Server/Client Communication Architecture ### Overview The MCP system consists of: 1. **Server**: Handles API requests and tool execution 2. **Client**: Communicates with both the server and LLM providers 3. **LLM Providers**: Process natural language queries and generate tool calls 4. **OpenAPI Parser**: Converts API specifications into executable tools ### Communication Flow 1. **Client Initialization**: - The client connects to the MCP server via SSE (Server-Sent Events) - It initializes an LLM provider (OpenAI or compatible) 2. **Query Processing**: - When a user sends a query, the client identifies the intent - Based on the intent, it selects appropriate tools to provide to the LLM 3. **LLM Interaction**: - The client sends the query and available tools to the LLM - The LLM generates a response, potentially including tool calls 4. **Tool Execution**: - When the LLM calls a tool, the client forwards the request to the MCP server - The server executes the corresponding API call - Results are returned to the client, which then sends them back to the LLM 5. **Response Generation**: - The LLM processes the tool results and generates a final response - The client returns this response to the user ### Key Components #### Server (server.py) - Uses FastMCP to handle client connections - Dynamically generates tools from OpenAPI specifications - Registers tools that map to API endpoints - Supports SSE transport for real-time communication #### Client (client.py) - Connects to the MCP server via SSE - Manages communication with LLM providers - Processes user queries and identifies intent - Handles tool calls from the LLM and forwards them to the server - Manages the conversation flow between user, LLM, and tools #### LLM Providers (openai_compatible.py) - Provides a unified interface to various LLM services - Handles chat completion requests with tool support - Manages API keys and authentication #### API Execution (http.py) - Handles actual HTTP requests to the Crawlab API - Manages authentication and token handling - Processes API responses