# Overview

Workflows combine API calls, AI processing, messages, and conditional logic in a sequence of steps. Send the steps and their input data to the [execution endpoint](/docs/execute-workflows) to run them.

## What is a workflow?

Each step has a `type` that selects its action and a `config` object that supplies its settings. An optional `output` name stores the step's result in context so later steps can use it.

## Anatomy of a workflow

* Steps run in order within each sequence. Use Batch for independent operations that can run in parallel.
* [Context](/docs/context-manipulation) contains the input data and named step results. Nested steps have their own scope.
* [Morpher templates](/docs/morpher) read context values in step configuration, for example `{{ user.email }}`.

## Available step types

| Step | Use it to |
| --- | --- |
| [Request](/docs/request) | Call an HTTP API |
| [Agent](/docs/agent) | Generate structured data with an LLM |
| [SMS](/docs/sms) and [Email](/docs/email) | Send messages through the configured providers |
| [Rule](/docs/rules) | Run the first branch whose conditions match |
| [Batch](/docs/batch) | Run independent steps in parallel |
| [Map](/docs/map) | Run a sequence for each item in an array |
| [Respond](/docs/respond) | Return a custom HTTP response and stop execution |

## Use cases

* Send email and SMS notifications after an order or booking.
* Extract structured fields from text with an Agent, then send the result to an API.
* Fetch data from several services and return a combined response.
* Route onboarding or alerts based on user data and conditions.

## Benefits

Steps share a configuration format and pass data through context, so you can combine them without writing a separate integration for each sequence. Start with the [execution guide](/docs/execute-workflows), then use the step references to configure each action.
