Understand the form
Fields, labels, choices, and constraints, returned as structured data your application can use.
THE ACTION LAYER FOR AI
Intelligence,
put to work.
The next frontier is doing.
We’re building the infrastructure that lets AI act on the web. Starting with forms.
Purpose-built intelligence. API + MCP.
SETTLEON / 001INTENT → ACTION
“Complete this form.”INTENT RECEIVED
Values verified.READY FOR YOUR NEXT STEP
FROM UNDERSTANDING TO EXECUTION 01—05
01 / THE PLATFORM
One interface.
A world of workflows.
The web was built
for people.
Open it to agents.
Every application, booking, and onboarding flow has a form between intent and outcome.
SettleOn is building a common interface for those interactions. Read the fields. Supply the values. Verify what changed. Submit when authorized.
Meet the API and MCP server ↗Act with precision
Write values and work through controls with an engine built specifically for form interactions.
Check the outcome
Read each value back. Report what is confirmed, what failed, and what remains uncertain.
02 / FROM REQUEST TO RESULT
INTERACTIVE PRODUCT PREVIEWIntent in.
Action out.
A shared language for the forms your agents encounter.
Explore an illustrative workflow below.
NNorthstar workspaceACCOUNT SETUP
A good place to start.
Let’s get your workspace ready.
Full name
Alex Morgan✓
Work email
alex@example.org✓
Team size
11–50 people✓
Create workspace→
Submission stays in your control.
SettleOn engineREADY
YOUR AGENT’S INTENT
“Set up Alex’s workspace with the supplied account details. Fill only.”
- 01Read✓
Identify fields and constraints.
- 02Fill✓
Apply the supplied values.
- 03Verify✓
Read back what the form holds.
3 fields confirmed · nothing submitted
This is an interactive illustration of the intended experience, not a connection to the live engine. PDF support is in development.
03 / BUILT FOR BUILDERS
Your agent.
Your stack.
Our action layer.
Connect from your application through HTTP and JSON, or give your AI assistant form tools through MCP. Both are designed around the same engine.
↗ Persistent form sessions↗ Structured field results↗ Explicit submission permissions
Get involved early ↗tools / form-interaction
// Tools for your AI assistant
open_form → Start a session
read_form → Understand the fields
list_choices → Inspect available options
fill_fields → Write and verify values
press_button → Act with permission
read_receipt → Inspect the result
close_form → End the sessionProposed interface. Endpoints and tool names may change before release. Public access is not open yet.
04 / THE WAY WE BUILD
Capability needs control.Built to act.
Designed to be accountable.
Purpose-built intelligence.
Our own specialized model and execution engine. Designed to handle forms without passing their contents to an external AI model.
Evidence with every action.
Every field gets an explicit result. An uncertain outcome stays uncertain. Failed interactions are reported as failures.
You set the boundaries.
Submission requires permission. Identity checks and signatures stay with the person. Your agent supplies the answers.
IN PRIVATE DEVELOPMENT
Build what
comes next.
We’re starting with forms.
The ambition is a more capable web.
We’re developing the engine, API, and MCP server with real workflows in mind. Bring us the forms standing between your agents and their next task.
NOW
Engine training and product development.
Early-access requests are open.