Learn how to connect Metorial's 600+ integrations to your Vercel AI SDK applications.

Basic Integration Pattern

The core pattern for using Metorial with AI SDK involves three steps:

1. Initialize the Metorial client
2. Create a provider session with your deployments
3. Pass the session tools to your AI SDK calls

Complete Example

import { metorialAISDK } from '@metorial/ai-sdk';
import { Metorial } from 'metorial';
import { generateText, streamText } from 'ai';
import { openai } from '@ai-sdk/openai';

const metorial = new Metorial({
  apiKey: process.env.METORIAL_API_KEY
});

await metorial.withProviderSession(
  metorialAISDK,
  { serverDeployments: ['your-deployment-id'] },
  async (session) => {
    // Use with generateText
    const result = await generateText({
      model: openai('gpt-4'),
      messages: [
        { role: 'user', content: 'What meetings do I have today?' }
      ],
      tools: session.tools,
      maxSteps: 10  // Allow multiple tool calls
    });
    
    console.log(result.text);
  }
);

Streaming Responses

Metorial works seamlessly with streaming:

await metorial.withProviderSession(
  metorialAISDK,
  { serverDeployments: ['your-deployment-id'] },
  async (session) => {
    const result = streamText({
      model: openai('gpt-4'),
      messages: [
        { role: 'user', content: 'Send a Slack message to #general' }
      ],
      tools: session.tools
    });
    
    for await (const chunk of result.textStream) {
      process.stdout.write(chunk);
    }
  }
);

Tool Execution

The AI SDK automatically handles tool execution when you provide tools. The model will:

1. Analyze available tools from your Metorial deployment
2. Decide which tools to call based on the user's request
3. Execute tools through Metorial's MCP infrastructure
4. Use the results to formulate its response

Session Lifecycle

The withProviderSession method manages the connection lifecycle:

• Opens a session when called
• Makes tools available during the callback
• Automatically closes the session when done
• Handles errors and cleanup

Error Handling

Always wrap your Metorial calls in try-catch blocks:

try {
  await metorial.withProviderSession(
    metorialAISDK,
    { serverDeployments: ['your-deployment-id'] },
    async (session) => {
      // Your code here
    }
  );
} catch (error) {
  console.error('Metorial session error:', error);
}

Common Use Cases

• Calendar management: "What's on my calendar this week?"
• Communication: "Send a message in Slack"
• Data retrieval: "Get the latest data from our database"
• Document management: "Summarize the document in Google Drive"
• Task management: "Create a Jira ticket for this bug"