Workflows Overview

​

Key Concepts

• Steps: Individual units of work with defined inputs and outputs
• Control Flow: Patterns like sequential (.then()), parallel (.parallel()), and conditional (.branch()) execution
• State Management: Shared state across all steps in a workflow
• Suspend/Resume: Human-in-the-loop capabilities for pausing and resuming execution
• Type Safety: Full TypeScript support with schema validation using Zod

​

Basic Workflow Structure

import { createWorkflow, createStep } from '@mastra/core';
import { z } from 'zod';

const myWorkflow = createWorkflow({
  id: 'my-workflow',
  inputSchema: z.object({
    input: z.string(),
  }),
  outputSchema: z.object({
    result: z.string(),
  }),
})
  .then(step1)
  .then(step2)
  .commit();

​

Graph-Based Execution

• Non-linear flows: Branch based on conditions, execute steps in parallel
• Dynamic routing: Choose execution paths at runtime
• State persistence: Save and restore workflow state at any point
• Time travel: Resume from any previous step in the workflow

​

Workflow Lifecycle

1. Define: Create workflow with input/output schemas and steps
2. Commit: Finalize the workflow graph structure
3. Execute: Run the workflow with input data
4. Monitor: Track step-by-step execution progress
5. Resume: Continue suspended workflows with additional data

// Define
const workflow = createWorkflow({ ... })
  .then(step1)
  .commit();

// Execute
const run = await workflow.execute({
  inputData: { input: 'Hello' },
});

// Monitor
const result = await run.result();
console.log(result.status); // 'success' | 'failed' | 'suspended'

​

Features

​

Type-Safe Schema Validation

​

Execution Context

• inputData: Data passed to this step
• state: Shared workflow state
• setState: Update workflow state
• getStepResult(): Access outputs from previous steps
• mastra: Access to the Mastra instance
• requestContext: Request-scoped data

​

Observability

• Distributed tracing support
• Step-level logging
• Error tracking
• Performance metrics

​

Next Steps