Implementation:Microsoft Autogen Studio MCP API
| Sources | python/packages/autogen-studio/frontend/src/components/views/mcp/api.ts |
|---|---|
| Domains | Frontend, API Client, MCP Protocol |
| Last Updated | 2026-02-11 |
Overview
McpAPI is a TypeScript API client class that provides comprehensive methods for managing MCP (Model Context Protocol) servers, galleries, resources, prompts, tools, and WebSocket connections in AutoGen Studio.
Description
The McpAPI class extends BaseAPI and implements a complete set of REST API methods for interacting with MCP servers. It provides functionality for:
- Gallery Management: List and retrieve galleries containing MCP workbenches
- MCP Server Operations: List/get resources, prompts, tools, and server capabilities
- Tool Execution: Call MCP tools with arguments and get results
- Connection Management: Test server connections, health checks, and WebSocket setup
- Workbench Extraction: Helper method to filter MCP workbenches from gallery components
The API client handles request/response serialization, error handling, and HTTP communication with the AutoGen Studio backend. It supports multiple MCP server types (StdioServerParams, SseServerParams, StreamableHttpServerParams) and provides typed interfaces for all data structures following the native MCP protocol specification.
Usage
The McpAPI is instantiated and used throughout the MCP view components to communicate with the backend. It's typically imported as a singleton instance (mcpAPI) and called with user credentials and server parameters.
Code Reference
Source Location: python/packages/autogen-studio/frontend/src/components/views/mcp/api.ts
Signature:
export class McpAPI extends BaseAPI {
// Gallery management
async listGalleries(userId: string): Promise<Gallery[]>
async getGallery(galleryId: number, userId: string): Promise<Gallery>
extractMcpWorkbenches(gallery: Gallery): Component<McpWorkbenchConfig>[]
// MCP Server operations
async listResources(serverParams: McpServerParams)
async getResource(serverParams: McpServerParams, uri: string)
async listPrompts(serverParams: McpServerParams)
async getPrompt(serverParams: McpServerParams, name: string, promptArgs?: Record<string, any>)
async getCapabilities(serverParams: McpServerParams): Promise<GetCapabilitiesResponse>
async listTools(serverParams: McpServerParams): Promise<ListToolsResponse>
async callTool(serverParams: McpServerParams, toolName: string, toolArguments: Record<string, any>): Promise<CallToolResponse>
// Connection management
async healthCheck(): Promise<{ status: boolean; message: string }>
async testMcpConnection(workbench: Component<McpWorkbenchConfig>): Promise<boolean>
async createWebSocketConnection(serverParams: McpServerParams): Promise<any>
}
Import:
import { mcpAPI } from './api';
I/O Contract
Inputs
| Method | Parameter | Type | Description |
|---|---|---|---|
| listGalleries | userId | string | User identifier for authentication |
| getGallery | galleryId | number | Gallery ID to retrieve |
| getGallery | userId | string | User identifier for authentication |
| listResources | serverParams | McpServerParams | Server connection parameters |
| getResource | serverParams | McpServerParams | Server connection parameters |
| getResource | uri | string | Resource URI to fetch |
| listPrompts | serverParams | McpServerParams | Server connection parameters |
| getPrompt | serverParams | McpServerParams | Server connection parameters |
| getPrompt | name | string | Prompt name to retrieve |
| getPrompt | promptArgs | Record<string, any> | Optional prompt arguments |
| listTools | serverParams | McpServerParams | Server connection parameters |
| callTool | serverParams | McpServerParams | Server connection parameters |
| callTool | toolName | string | Name of tool to call |
| callTool | toolArguments | Record<string, any> | Arguments for tool execution |
| testMcpConnection | workbench | Component<McpWorkbenchConfig> | Workbench to test |
Outputs
| Method | Return Type | Description |
|---|---|---|
| listGalleries | Promise<Gallery[]> | Array of available galleries |
| getGallery | Promise<Gallery> | Single gallery object with full configuration |
| extractMcpWorkbenches | Component<McpWorkbenchConfig>[] | Filtered array of MCP workbenches |
| listResources | Promise<any> | List of available MCP resources |
| getResource | Promise<any> | Resource content and metadata |
| listPrompts | Promise<any> | List of available prompts |
| getPrompt | Promise<any> | Prompt content and metadata |
| getCapabilities | Promise<GetCapabilitiesResponse> | Server capabilities object |
| listTools | Promise<ListToolsResponse> | List of available tools with schemas |
| callTool | Promise<CallToolResponse> | Tool execution result |
| healthCheck | Promise<{status: boolean, message: string}> | Health status response |
| testMcpConnection | Promise<boolean> | Connection test success status |
| createWebSocketConnection | Promise<any> | WebSocket connection details |
Usage Examples
Listing galleries for a user:
const galleries = await mcpAPI.listGalleries(user.id);
Extracting MCP workbenches from a gallery:
const gallery = await mcpAPI.getGallery(galleryId, userId);
const workbenches = mcpAPI.extractMcpWorkbenches(gallery);
Listing tools from an MCP server:
const toolsResponse = await mcpAPI.listTools(serverParams);
if (toolsResponse.status && toolsResponse.tools) {
console.log('Available tools:', toolsResponse.tools);
}
Calling an MCP tool:
const result = await mcpAPI.callTool(
serverParams,
'get_weather',
{ location: 'San Francisco' }
);
Testing server connection:
const isConnected = await mcpAPI.testMcpConnection(workbench);
if (isConnected) {
console.log('Server is available');
}