Workflows let your team build multi-step automation sequences that run AI tasks automatically. Trigger them on demand, from an external system via webhook, or on a recurring schedule. Add approval gates when you need a human to review before a step proceeds.

​
Creating a workflow

1

Open Workflows

Navigate to Workflows in the sidebar and click New Workflow.
2

Define the steps

Build your workflow by adding steps. Each step can call an AI model, transform data, call an external API, or wait for approval. Connect steps in sequence or add conditional branching.
3

Choose a trigger

Select how the workflow starts: manually on demand, via an incoming webhook from an external system, or on a cron schedule.
4

Save and run

Save the workflow. Run it manually from the workflow detail page to test it before enabling the automatic trigger.

​
Trigger types

Run the workflow on demand by clicking Run Now on the workflow detail page. Useful for tasks you want to initiate yourself, or for testing before automating.

​
Approval gates

Add an approval gate to any step to pause the workflow and wait for a human to review before continuing. When a workflow reaches an approval gate:
  1. An item appears in the Inbox of the designated approvers
  2. Approvers open the inbox item, review the workflow’s progress so far, and click Approve or Reject
  3. If approved, the workflow resumes from the next step. If rejected, the run is stopped and marked as rejected.

​
Viewing run history

Every workflow run is logged with a full record of each step — inputs, outputs, timing, and any errors. Open any workflow and click the Runs tab to see the history. Click into any individual run to inspect it step by step.

​
Inbox

Approval requests and workflow notifications appear in your Inbox (the bell icon in the top navigation). The inbox shows unread items with a badge count. Click any item to open the relevant workflow run and take action.

​
Real-time updates

Workflow run completions and inbox items appear in real time — you don’t need to refresh the page. When ManyLayers is deployed with Redis, real-time events are delivered across all server replicas so every connected user receives updates regardless of which instance they are connected to.