MGBox
From submission to bind
One path through the MGA, with the authority check built into the path rather than bolted onto it. Every step leaves a record, so what was decided and who was allowed to decide it is answerable later without reconstructing it from email.
The submission arrives
An appointed broker submits once. The application and exposure schedule land as structured data, not as an attachment somebody has to retype, and the file opens with the public-record check already run.
It is priced against the ratebook
The rating plan is a versioned artifact rather than a spreadsheet on somebody's desktop. Every quote records which ratebook version produced it, so a number can be explained months later.
Authority decides who answers
Limits, appetite and referral triggers are configuration. Inside authority an underwriter quotes; outside it the file routes to whoever does hold that authority, with the reason attached. Nobody has to remember where the line is.
Bind, and the record closes
Binding writes the policy, the schedule and the elections together. Bordereau and regulatory reporting are produced from that record rather than assembled afterwards from three systems.
MGBox is not offered publicly. We run our own MGA on it.
Why we built it
The plan is not to sell software. It is to find out whether an MGA can be run this way at all, on our own book, before anyone else has to trust it.
Tenant zero
We are the first and hardest user. Every gap in the platform is a gap in our own underwriting day, which is a considerably faster feedback loop than a customer filing a ticket.
A new line is a workflow, not a rebuild
Products live in a registry: agreements, elections, rating inputs and authority rules as data. Launching a line is configuring one, which is what makes a small team able to carry several.
The record is the point
An AI-native MGA only works if every automated step is attributable and reversible. The platform is built so that a decision, a price and an authority can each be traced back to their inputs.