Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.sequentum.com/llms.txt

Use this file to discover all available pages before exploring further.

Agent Builder turns natural language into a working extraction Agent. You describe the data you want; Agent Builder plans, navigates, parses, and validates — then hands you a real, versioned Agent you can run, schedule, or share. This is the default authoring path in Sequentum Cloud. From here, you’ll typically run the Agent via the API. Reach for the Editor only when Agent Builder can’t yet handle a corner case (custom scripts, complex navigation, niche commands).

How it works

1

Describe

“Pull product name, price, and SKU from https://example-store.com/products/.”
2

Plan

Agent Builder loads the site, identifies the structure, and proposes a command tree.
3

Test

The plan is executed against the live site. Agent Builder reports back: pages visited, records extracted, anything blocked.
4

Iterate

Send a follow-up — “include image URLs” or “skip out-of-stock items” — and Agent Builder revises.
5

Finish

Save the session as a real Agent. From here it behaves like any other Agent: runnable, schedulable, versionable.

Two ways to use it

Control Center UI

Click “New agent → Build with AI.” Best for one-off agents and exploration.

API

Build agents programmatically. Best when agent creation is part of a pipeline or product feature you ship to your own users.

Session lifecycle

A Builder session is a stateful object with five statuses — processing, ready, completed, error, cancelled. You start a session with a prompt, watch events stream as Agent Builder plans and tests, and end with completed (Agent saved) or cancelled. At launch the experience is single-shot — describe the agent comprehensively in your first prompt and it’s built and ready to run. See Sessions for the full state machine.

What’s next