> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.meetgail.com/platform/workflows/getting-started/how-workflows-run/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.meetgail.com/_mcp/server.

# How Workflows Run

> The trigger, steps, and end model behind every workflow, plus a tour of the builder canvas.

Every workflow follows the same simple shape: something kicks it off, a series of
steps run in order, and then it finishes. Once you can picture that flow, the
builder canvas makes sense at a glance.

## The shape of a run

A workflow always moves in one direction, from a beginning to an end:

```
  Start   ->   Steps   ->   End
 (kickoff)   (the work)   (finish)
```

* **Start** is where a run begins. It is the single entry point of the workflow,
  and it can capture information to work with, such as a new lead's name and phone
  number. See [Starting a Workflow](/platform/workflows/core-concepts/starting-a-workflow).
* **Steps** are the nodes in the middle that do the actual work: sending a message,
  adding a contact, making a decision, and so on. They run one after another,
  following the connections you draw.
* **End** marks where a run stops. When the workflow reaches an End step, that run
  is complete.

Each run flows from Start, through the connected steps, to End. If you run the same
workflow ten times, you get ten separate runs, each following the same path with
its own data.

## What makes a run begin

A workflow does not run on its own. It starts when it is triggered. A run can
begin because someone starts it by hand, because another system sends it a signal,
or on a schedule you set. You can read more in
[Triggers & Webhooks](/platform/workflows/core-concepts/triggers-webhooks).

## How data moves between steps

Steps are not isolated. Each step can use information produced by the steps before
it. The Start step might capture a phone number, and a later Send SMS step can text
that exact number. This passing of data from one step to the next is what lets a
workflow react to real information instead of fixed values. See
[Variables & Data Flow](/platform/workflows/core-concepts/variables-data-flow).

## When a step has trouble

Steps are built to be reliable. If a step hits a temporary problem, the workflow
gives it another chance or two before giving up, and some steps that should never
run twice are set to run only once. You do not have to configure any of this to get
started. For the details, see
[Timeouts & Retries](/platform/workflows/core-concepts/timeouts-retries).

## A tour of the canvas

You build and read workflows on a visual canvas. Each step is a node, and the
lines between the nodes are the connections that set the order of the run.

![The builder canvas showing a workflow of four connected steps: Start, Send SMS, Add Contact, and End.](/_fern-img/d193c9a54de5639706752f5e60d7fa49be98a71d1d7abbb78c496bb4fa9c7eeb.webp)

Reading the canvas above, top to bottom:

* The **Start** node at the top is the entry point.
* Each **step** below it is a node that does one job. Selecting a node opens its
  settings so you can configure it.
* The **connections** between nodes are the arrows that define the order steps run
  in.
* The **End** node at the bottom finishes the run.

> **Note**
>
> If the canvas ever feels crowded, use Fit View to bring every step back into view
> at once.

## Where to go next

#### [Build Your First Workflow](/platform/workflows/getting-started/build-your-first-workflow)

Put this model to work and build a real workflow step by step.

#### [Nodes & Connections](/platform/workflows/core-concepts/nodes-connections)

Go deeper on how steps connect and pass control to one another.

#### [Starting a Workflow](/platform/workflows/core-concepts/starting-a-workflow)

Learn what the Start step can capture when a run begins.

#### [Testing & Debugging](/platform/workflows/core-concepts/testing-debugging)

Watch a run happen step by step and see what each one did.

> Documentation for Gail, the AI platform for financial services. Learn how to set up GailGPT and Gail Agent to automate customer communications for insurance, banking, and finance.