With Checkpoints, labeled evidence will be captured at critical decision points, which will allow you to quickly verify correctness and have clear visibility into what the automation accomplished.
Setting a Checkpoint ID
Before you begin: Determine your Checkpoint ID. This will be the unique "unit" or "item" processed and followed throughout the workflow. For example, a patient ID number.
You may set only ONE checkpoint ID in an automation, but a unique value will be returned every time the agent runs. Each checkpoint ID value will appear as a row in the checkpoints table.
1. Navigate to an automation
2. Go to a new line and type " /"
3. Scroll through the list of available Commands, or start typing to search for "setCheckpointId" then click to select
4. Fill in the form fields to customize your Checkpoint ID (refer to the "Raw preview" to understand how this will be translated into instructions for the agent)
5. Click "Done"
Adding Checkpoints
Before you begin: Determine your checkpoints. These will be any meaningful outcomes, decisions, or states in the automation that you want to highlight. For example, "insurance eligibility status determined" or "appointment confirmation sent."
Each checkpoint will appear as a column in the checkpoints table.
Tip: Place your Checkpoint right before a destructive action (e.g. clicking submit, send, or next) or key decision is made (e.g. marking a claim as ineligible) so that the screenshot displays the information needed in order to verify correctness.
For example, if your agent is meant to fill in a prior authorization form, you’ll want to insert a checkpoint after the last field is filled (✅ the completed form with patient name, DOB, SSN, etc. is visible in the screenshot) and before the form is submitted (❌ once submitted, only the generic confirmation page is visible in the screenshot).
6. Navigate to an agent
7. Go to a new line where you'd like your Checkpoint to be and type " /"
8. Scroll through the list of available Commands, or start typing to search for "Checkpoint" then click to select
9. Fill in the form fields to customize your Checkpoint
Name: A short title for the outcome/decision point you're checking (this will appear as the column header)
Outcome: Any input(s), value(s) on the screen, or static text that will act as supporting evidence.
Screenshot: Optionally include a screenshot of what the agent is seeing on the screen as supporting evidence.
10. Click "Done"
11. Use the "Add checkpoint" button on an agent if you wish to capture the agent’s complete list of inputs and outputs as evidence
12. Give the Checkpoint a name: A short title for the outcome/decision point you're checking (this will appear as the column header)
13. Optionally include a screenshot of what the agent is seeing on the screen as supporting evidence
14. Click "Add"
15. When you're finished setting up your Checkpoints, click "Save"
