Workflows

Automate multi-step business processes triggered by chat events, schedules, or external messages.

What is a Workflow?

A workflow is a visual automation that runs a sequence of steps — called nodes — when triggered. Each node does one thing: call an AI agent, send an email, update a data table, or send a WhatsApp message. Nodes are connected in a graph and executed in order.

Workflows can run automatically (triggered by a chat pattern, an incoming email, or a schedule) or manually from the admin portal.

Trigger Types

Every workflow has exactly one trigger that determines when it runs.

Trigger	How it fires	Typical use case
`manual`	Fired by clicking 'Execute' in the admin portal or via webhook.	Testing, one-off operations, admin-initiated tasks.
`cron`	Fired on a schedule using a cron expression.	Daily reports, nightly data sync, weekly summaries.
`webhook`	Fired by a POST request to the workflow's webhook URL.	Connecting external systems — e-commerce, CRM, payment providers.
`email`	Fired when a new email arrives in a connected mailbox.	Auto-reply to enquiries, routing support emails, summarising inboxes.
`whatsapp`	Fired when a WhatsApp message is received on a connected number.	Automated WhatsApp replies, order confirmations, appointment reminders.
`chat_pattern`	Fired when a chat message matches a configured pattern.	Escalation workflows, lead capture when intent is detected.

Node Types

Each node in the workflow graph performs one action.

Agent Call

Sends a prompt to an AI agent and stores the response. Use this to generate replies, summaries, or decisions within the workflow.

Send Email

Sends an email via a connected mailbox. Supports dynamic content using data from previous nodes.

Send WhatsApp

Sends a WhatsApp message via a connected Meta or Twilio connector. Great for notifications and follow-ups.

Data Table Operations

Query, insert, update, upsert, or delete rows in any of your data tables. Results are available to subsequent nodes.

Using the Visual Editor

Go to Workflows → your workflow → Editor to open the drag-and-drop canvas.

Click Add node and choose a node type from the panel.
Configure the node — fill in templates, select agents or tables, map input variables.
Drag from one node's output handle to another node's input to connect them.
Nodes execute top-to-bottom, left-to-right. Outputs from one node are available as variables in the next.
Click Save graph, then Execute to test. Check the Sessions tab to see the run log.

Using a Workflow as an AI Tool

When you enable <em>Use as AI tool</em> on a workflow, agents can invoke it during conversations. The agent decides when to call the tool based on the tool description you provide.

Example: A workflow that queries a pricing data table is marked as a tool with description 'Get current prices for products'. When a visitor asks 'How much does the Enterprise plan cost?', the agent calls the workflow, gets the price, and answers accurately.

Workflow Sessions (Run Log)

Every workflow execution creates a session record. View it under Workflows → Sessions to see exactly what happened at each node.

Status	Meaning
success	All nodes completed without error.
failed	One or more nodes encountered an error. Check the session detail for the error message and which node failed.
running	Workflow is currently executing. Refresh to see the updated status.

Tip: Start with a manual trigger while building and testing. Switch to your real trigger (cron, email, whatsapp) only after confirming the workflow behaves correctly.