Checkpoint
The Checkpoint block pauses a workflow at a defined point and presents the user with a prompt to review the current state before the workflow continues.
Pause a workflow for user review or revision
The Checkpoint block pauses a workflow at a defined point and presents the user with a prompt to review the current state before the workflow continues. It supports two modes: Alert mode, which asks the user to approve or reject and then routes the workflow accordingly, and Revise Variable mode, which lets the user interact with an AI model to edit a specific variable — either plain text or HTML — before proceeding.
Key configuration fields include a title and optional description shown to the user, labels for the approve and reject actions (defaulting to "Approve" and "Reject"), and separate transition targets for approval and rejection paths. In Revise Variable mode, you additionally specify which variable the user will edit, a label for that variable, its type (text or HTML), optional dimensions for HTML previews, and guidance text that helps the AI model understand how to handle the user's revision requests.
This block fits into workflows where human oversight or collaboration is needed at an intermediate step — for example, reviewing AI-generated content before it is sent, confirming a decision before a consequential action is taken, or iteratively refining a document with AI assistance before moving to the next stage.
What you can build
Real-world workflows powered by the Checkpoint block.
Approve AI-Generated Email
Pause a workflow after an AI drafts an outbound email so the user can review and approve or reject the message before it is sent.
Refine a Written Report
Let a user collaborate with an AI model to revise a generated text report before it is saved or distributed.
Review HTML Document Output
Present a rendered HTML preview to the user and allow them to request AI-assisted edits before the document is finalized.
Confirm Data Before Submission
Insert a checkpoint before a form or database write so the user can verify the collected data and reject the flow if something looks wrong.
Gate a Multi-Step Process
Use an alert-mode checkpoint to require explicit user confirmation before a workflow advances to a sensitive or irreversible step.
Iterate on a Product Spec
Allow a user to chat with an AI model to refine a product specification variable through multiple rounds of revision before the workflow proceeds.
Ready to add Checkpoint to your workflow?
Get Started FreeCommon questions about Checkpoint
What are the two modes available in the Checkpoint block?
The block supports Alert mode and Revise Variable mode. Alert mode presents the user with an approve/reject prompt and routes the workflow to different steps based on their choice. Revise Variable mode lets the user work with an AI model to edit a designated variable (either text or HTML) before the workflow continues.
What parameters are required to configure this block?
Required fields include a title for the checkpoint, an approve transition target (approveStepId), and the mode (alert or reviseVariable). A description is optional. In Revise Variable mode, you must also provide the variable name, a variable label, the variable type (text or HTML), a revision mode (chat), and revision guidance for the AI model.
What does the Checkpoint block return as output?
The block does not define a structured output payload. Its primary effect is controlling workflow routing — directing execution to the approve path or the reject path based on the user's response. In Revise Variable mode, the specified variable is updated with the user's revised content.
Can the approve and reject button labels be customized?
Yes. The approveLabel and rejectLabel fields allow custom button text. Both fields are optional and default to "Approve" and "Reject" respectively if left unset.
What kinds of workflows commonly use the Checkpoint block?
It is used in workflows that require human review before a consequential action, such as sending a message, writing to a database, or publishing content. It is also used in document or content creation workflows where a user needs to iteratively refine AI-generated text or HTML before the workflow moves forward.
Related capabilities
Start
Specifies how the workflow is invoked
User Input
Prompt the user for one or more pieces of data that will be saved as a variable. Can include text, multiple choice, images, or files.
Run Function
Execute a custom JavaScript or Python function
Chat
Start a chat session with an LLM.
End
End the workflow. Optionally return a structured output (for API) or send notifications.
Jump
Transition to another workflow, maintaining the same scope
Add Checkpoint to your workflow
Build powerful AI workflows with drag-and-drop blocks. No coding required.