Basics

Simple agent with tables

This agent receives a support ticket and categorizes it into a database table.

Prompt template (categorize_ticket.md):

You received a support ticket:

- From: {{ trigger_task.payload.customer_email }}
- Subject: {{ trigger_task.payload.subject }}
- Message: {{ trigger_task.payload.message }}

Categorize this ticket by inserting a row into the `support_tickets` table with:
- email: the customer's email
- category: one of "billing", "technical", "general"
- priority: one of "low", "medium", "high"
- summary: a one-sentence summary of the issue

Permissions: tables_insert_support_tickets

The agent reads the ticket, decides the category and priority, and inserts a row — all autonomously.

Agent with connections

This agent sends a Slack message when a new order is placed.

Prompt template (notify_order.md):

A new order was placed:

- Customer: {{ trigger_task.payload.customer_name }}
- Total: ${{ trigger_task.payload.total }}
- Items: {{ trigger_task.payload.items | length }} items

Send a message to the #orders Slack channel with a summary of this order.

Permissions: Slack connection with send message action enabled.

The agent searches for the Slack send message action, then executes it with the order details.

Agent with workflow transitions

This agent reviews a document and routes it to the appropriate next stage.

Prompt template (review_document.md):

Review the following document submission:

- Type: {{ trigger_task.payload.document_type }}
- Submitted by: {{ trigger_task.payload.submitted_by }}

The document is stored at: {{ trigger_task.payload.file_path }}

Read the document and determine if it is complete. Then:
- If complete, send a task of type "approved" with the document summary
- If incomplete, send a task of type "rejected" with the missing fields

Permissions: File read access.

Workflow transitions: approved and rejected task types, each connecting to a different downstream stage.

Agent with images

This agent analyzes an image attached to the task.

Prompt template (analyze_receipt.md):

The user submitted this receipt:

{{ trigger_task.payload.receipt_image }}

Extract the following information:
- Store name
- Date
- Total amount
- List of items with prices

Insert the receipt data into the `receipts` table.

Permissions: tables_insert_receipts

The image is automatically sent to the LLM as visual content — the agent sees it directly and extracts the data without needing any special tool.