Skip to main content
When you execute a Blueprint, Blueprint runs each node in sequence, passing the output of each step as the input to the next. This page explains how to trigger an execution, what happens during the run, how to monitor progress, and how to handle nodes that require human input before the chain can continue.

​
How execution works

Blueprint execution is always sequential. Nodes run one at a time in the order they appear in the chain:
  1. The execution engine resolves all {{variable}} references in the first node’s configuration.
  2. The first node runs and produces an output.
  3. The output is stored and made available to the next node as its input.
  4. Steps 1–3 repeat for every subsequent node.
  5. When the last node completes, the execution is marked as completed.
If any node fails and is not handled by a conditional node, the execution stops immediately with a failed status.

​
Execution statuses

​
Triggering an execution

  • Web UI
  • API
  • CLI
1 Open the Blueprint Navigate to your Blueprint from the dashboard and open it in the editor. 2 Click Run Click the Run button in the top-right toolbar. If the Blueprint declares input variables, a dialog appears asking you to fill them in. 3 Supply inputs and execute Fill in any required variable values, then click Execute. The execution panel opens automatically and shows live progress. Call the execute procedure with your Blueprint ID and any input variables:
The response includes an executionId you can use to poll for status:
Use the iblueprint execute command with your Blueprint ID and a JSON inputs string:
The CLI prints the execution ID when the run starts:
Check the status at any time using the returned execution ID:

​
Monitoring execution

​
Execution log

Every execution generates a step-by-step log you can access from the Executions tab on any Blueprint. The log shows:
  • The status of each node (pending, running, completed, failed, skipped)
  • The input passed to each node
  • The output produced by each node
  • Timing and token usage for LLM nodes
  • Error messages for failed nodes

​
Status polling (API)

If you triggered an execution via the API, poll the status procedure using the executionId:

​
Human-in-the-loop

When a Blueprint contains a human, human_checkpoint, or approval_gate node, execution pauses when that node is reached. The execution status changes to paused and Blueprint notifies the designated reviewer or approver. What happens during a pause 1 Execution pauses The chain stops at the human node. All previously completed nodes retain their outputs, and no downstream nodes run yet. 2 Reviewer is notified Blueprint sends a notification (email, in-app, or webhook, depending on your settings) to the person or role assigned in the node’s assignee configuration. 3 Reviewer submits input or approves The reviewer opens the paused execution from their dashboard, reviews the context, fills in any requested fields (for human nodes), and submits their response or approval decision. 4 Execution resumes The human node’s output (the reviewer’s response) is passed to the next node and the chain continues from where it left off.

​
Cancelling an execution

You can cancel a running or paused execution at any time from the Executions tab on your Blueprint by clicking Cancel. Cancelled executions retain their partial logs and the outputs of any nodes that completed before cancellation.