Skip to main content

​
What an automation does

An automation runs repeatable operational work. Describe what you need in plain language, and Flow turns it into an automation you can review. As you refine the work, Flow updates the draft, adds the inputs and integration actions the run needs, and flags issues to fix. You stay in control throughout: review the draft, run Validation scenarios, try simulated or live actions, and decide when to publish. Once published, an automation can start from a Flow request, the API, MCP, a schedule, an event, or another automation. Before building, define the outcome, participants, inputs, approvals, and systems involved. Flow asks for missing process details instead of assuming what an unclear request means.
You describe the work in plain language and review, test, and publish. Flow drafts the automation, adds inputs and integration actions, and flags issues. You stay in control from first draft to publish.You describe the work in plain language and review, test, and publish. Flow drafts the automation, adds inputs and integration actions, and flags issues. You stay in control from first draft to publish.

​
Core concepts

​
How automations move through Flow

  1. Describe the work. Tell Flow what work should happen and what information the run needs.
  2. Review the draft automation. Review the steps, inputs, and integration actions. Open the code only when you need implementation detail.
  3. Select integration access. Choose or request the provider connections needed for live tests and published runs.
  4. Review Validation scenarios and test the draft. Run Validation scenarios for the saved revision, then use simulated actions and live connections when you intend to call providers.
  5. Publish a version. Publish the automation for a team or the organization.
  6. Trigger published runs. Start the published version from a configured request, schedule, event, API, MCP, or parent automation.
  7. Monitor run history. Check each run’s status, version, provider calls, retries, and errors.

​
Choose where to start

First automation

Create a simple automation with Flow and test it in Simulated mode.

Build automations

Review automation anatomy, integration actions, Validation scenarios, and publishing.

Set up integrations

Authorize provider accounts and allowed actions for automation runs.

Trigger automations

Start published automations from any configured entry point.

Monitor automations

Inspect run status, versions, provider calls, retries, and errors.

​
Frequently asked questions

Yes. Define the onboarding steps, required information, approvals, and connected systems. Flow can draft the automation for you to review, test, and publish. It is not active until you publish it.
Yes. You can write automation code yourself, then use Flow to review, refine, or extend it before publishing.
Yes. Use Validation scenarios for saved revision checks, start interactive runs in Simulated mode, then switch to Live when you intend to call providers.
From any configured entry point, including a Flow request, schedule, event, API, MCP, webhook, or another automation.
Automations call providers through integration actions. Setting up integrations and authorizing provider accounts is covered in Set up automation integrations.
The same provider can support different Flow surfaces. A Google Drive connection for knowledge search does not need the same permissions as a Google Drive integration used by an automation.Automation integrations are action-scoped. Administrators choose which team can use a connection, which actions are allowed, and which provider permissions Flow requests. This keeps automation access separate from knowledge, asset, and identity integrations.