--- title: experimental_createMCPClient description: Create a client for connecting to MCP servers --- # `experimental_createMCPClient()` Creates a lightweight Model Context Protocol (MCP) client that connects to an MCP server. The client provides: - **Tools**: Automatic conversion between MCP tools and AI SDK tools - **Resources**: Methods to list, read, and discover resource templates from MCP servers - **Prompts**: Methods to list available prompts and retrieve prompt messages - **Elicitation**: Support for handling server requests for additional input during tool execution It currently does not support accepting notifications from an MCP server, and custom configuration of the client. This feature is experimental and may change or be removed in the future. ## Import ## API Signature ### Parameters Promise', description: 'A method that starts the transport', }, { name: 'send', type: '(message: JSONRPCMessage) => Promise', description: 'A method that sends a message through the transport', }, { name: 'close', type: '() => Promise', description: 'A method that closes the transport', }, { name: 'onclose', type: '() => void', description: 'A method that is called when the transport is closed', }, { name: 'onerror', type: '(error: Error) => void', description: 'A method that is called when the transport encounters an error', }, { name: 'onmessage', type: '(message: JSONRPCMessage) => void', description: 'A method that is called when the transport receives a message', }, ], }, { type: 'MCPTransportConfig', parameters: [ { name: 'type', type: "'sse' | 'http", description: 'Use Server-Sent Events for communication', }, { name: 'url', type: 'string', description: 'URL of the MCP server', }, { name: 'headers', type: 'Record', isOptional: true, description: 'Additional HTTP headers to be sent with requests.', }, { name: 'authProvider', type: 'OAuthClientProvider', isOptional: true, description: 'Optional OAuth provider for authorization to access protected remote MCP servers.', }, ], }, ], }, { name: 'name', type: 'string', isOptional: true, description: 'Client name. Defaults to "ai-sdk-mcp-client"', }, { name: 'onUncaughtError', type: '(error: unknown) => void', isOptional: true, description: 'Handler for uncaught errors', }, { name: 'capabilities', type: 'ClientCapabilities', isOptional: true, description: 'Optional client capabilities to advertise during initialization. For example, set { elicitation: {} } to enable handling elicitation requests from the server.', }, ], }, ], }, ]} /> ### Returns Returns a Promise that resolves to an `MCPClient` with the following methods: Promise>`, description: 'Gets the tools available from the MCP server.', properties: [ { type: 'options', parameters: [ { name: 'schemas', type: 'TOOL_SCHEMAS', isOptional: true, description: 'Schema definitions for compile-time type checking. When not provided, schemas are inferred from the server.', }, ], }, ], }, { name: 'listResources', type: `async (options?: { params?: PaginatedRequest['params']; options?: RequestOptions; }) => Promise`, description: 'Lists all available resources from the MCP server.', properties: [ { type: 'options', parameters: [ { name: 'params', type: "PaginatedRequest['params']", isOptional: true, description: 'Optional pagination parameters including cursor.', }, { name: 'options', type: 'RequestOptions', isOptional: true, description: 'Optional request options including signal and timeout.', }, ], }, ], }, { name: 'readResource', type: `async (args: { uri: string; options?: RequestOptions; }) => Promise`, description: 'Reads the contents of a specific resource by URI.', properties: [ { type: 'args', parameters: [ { name: 'uri', type: 'string', description: 'The URI of the resource to read.', }, { name: 'options', type: 'RequestOptions', isOptional: true, description: 'Optional request options including signal and timeout.', }, ], }, ], }, { name: 'listResourceTemplates', type: `async (options?: { options?: RequestOptions; }) => Promise`, description: 'Lists all available resource templates from the MCP server.', properties: [ { type: 'options', parameters: [ { name: 'options', type: 'RequestOptions', isOptional: true, description: 'Optional request options including signal and timeout.', }, ], }, ], }, { name: 'listPrompts', type: `async (options?: { params?: PaginatedRequest['params']; options?: RequestOptions; }) => Promise`, description: 'Lists available prompts from the MCP server.', properties: [ { type: 'options', parameters: [ { name: 'params', type: "PaginatedRequest['params']", isOptional: true, description: 'Optional pagination parameters including cursor.', }, { name: 'options', type: 'RequestOptions', isOptional: true, description: 'Optional request options including signal and timeout.', }, ], }, ], }, { name: 'getPrompt', type: `async (args: { name: string; arguments?: Record; options?: RequestOptions; }) => Promise`, description: 'Retrieves a prompt by name, optionally passing arguments.', properties: [ { type: 'args', parameters: [ { name: 'name', type: 'string', description: 'Prompt name to retrieve.', }, { name: 'arguments', type: 'Record', isOptional: true, description: 'Optional arguments to fill into the prompt.', }, { name: 'options', type: 'RequestOptions', isOptional: true, description: 'Optional request options including signal and timeout.', }, ], }, ], }, { name: 'onElicitationRequest', type: `( schema: typeof ElicitationRequestSchema, handler: (request: ElicitationRequest) => Promise | ElicitResult ) => void`, description: 'Registers a handler for elicitation requests from the MCP server. The handler receives requests when the server needs additional input during tool execution.', properties: [ { type: 'parameters', parameters: [ { name: 'schema', type: 'typeof ElicitationRequestSchema', description: 'The schema to validate requests against. Must be ElicitationRequestSchema.', }, { name: 'handler', type: '(request: ElicitationRequest) => Promise | ElicitResult', description: 'A function that handles the elicitation request. The request contains a message and requestedSchema. The handler must return an object with an action ("accept", "decline", or "cancel") and optionally content when accepting.', }, ], }, ], }, { name: 'close', type: 'async () => void', description: 'Closes the connection to the MCP server and cleans up resources.', }, ]} /> ## Example ```typescript import { experimental_createMCPClient as createMCPClient, generateText, } from '@ai-sdk/mcp'; import { Experimental_StdioMCPTransport } from '@ai-sdk/mcp/mcp-stdio'; import { openai } from '@ai-sdk/openai'; let client; try { client = await createMCPClient({ transport: new Experimental_StdioMCPTransport({ command: 'node server.js', }), }); const tools = await client.tools(); const response = await generateText({ model: 'anthropic/claude-sonnet-4.5', tools, messages: [{ role: 'user', content: 'Query the data' }], }); console.log(response); } catch (error) { console.error('Error:', error); } finally { // ensure the client is closed even if an error occurs if (client) { await client.close(); } } ``` ## Error Handling The client throws `MCPClientError` for: - Client initialization failures - Protocol version mismatches - Missing server capabilities - Connection failures For tool execution, errors are propagated as `CallToolError` errors. For unknown errors, the client exposes an `onUncaughtError` callback that can be used to manually log or handle errors that are not covered by known error types.