Learn how to build sophisticated agentic workflows using Metorial's integrations with the AI SDK.

What are Agentic Workflows?

Agentic workflows allow your AI to autonomously use multiple tools in sequence to accomplish complex tasks. With Metorial and the AI SDK, you can build agents that interact with hundreds of services seamlessly.

Basic Agentic Pattern

import { metorialAiSdk } from '@metorial/ai-sdk';
import { Metorial } from 'metorial';
import { openai } from '@ai-sdk/openai';
import { generateText } from 'ai';

const metorial = new Metorial({
  apiKey: 'your-metorial-api-key'
});

await metorial.withProviderSession(
  metorialAiSdk,
  { serverDeployments: ['your-server-deployment-id'] },
  async session => {
    const result = await generateText({
      model: openai('gpt-4'),
      tools: session.tools,
      maxSteps: 15, // Allow multiple tool calls
      prompt: `
        Analyze my productivity for this week:
        1. Check my calendar for meetings
        2. Review my GitHub commits
        3. Summarize my Slack messages
        4. Create a weekly report and email it to me
      `
    });
    
    console.log(result.text);
  }
);

Multi-Integration Workflows

Combine different integrations for powerful workflows:

const result = await generateText({
  model: openai('gpt-4'),
  tools: session.tools,
  maxSteps: 20,
  prompt: `
    Project kickoff workflow:
    1. Create a new GitHub repository for "Project Phoenix"
    2. Set up initial project structure
    3. Create a Slack channel for the team
    4. Schedule a kickoff meeting in Google Calendar
    5. Create tasks in our project management tool
    6. Send a summary email to all stakeholders
  `
});

Conditional Tool Usage

The AI SDK with Metorial tools can make intelligent decisions about which tools to use:

const result = await generateText({
  model: openai('gpt-4'),
  tools: session.tools,
  prompt: `
    If there are any high-priority emails from the CEO:
    - Create a calendar reminder
    - Post in the leadership Slack channel
    - Mark the email as important
    
    Otherwise, just give me a summary of today's emails.
  `
});

Streaming Agentic Workflows

Show progress in real-time:

import { streamText } from 'ai';

const result = streamText({
  model: openai('gpt-4'),
  tools: session.tools,
  maxSteps: 10,
  prompt: 'Research topic X, create a summary document, and share it with the team'
});

for await (const chunk of result.textStream) {
  process.stdout.write(chunk);
}

Best Practices

• Set appropriate maxSteps to prevent infinite loops
• Provide clear, specific prompts for better tool selection
• Use streaming for long-running workflows
• Handle errors gracefully
• Test workflows with different scenarios

With Metorial handling 600+ integrations, you can focus on designing intelligent agent behaviors rather than building and maintaining integration code.