A rules engine decides. The AI just explains.
A language model reads what you type. A rules engine — the same kind of logic real eligibility screeners use — decides what you qualify for. The model never gets a vote on your results; it only translates the outcome into plain English.
You describe it
No forms, no account
A rules engine decides
Same rules real screeners use
Plain-language answer
What to apply for, where, what to bring
Where this came from
OpenHand grew out of watching real mutual aid requests and noticing how often what people were asking their community for was something a government program was already meant to cover.
Alongside mutual aid, not instead of it
Mutual aid networks already do real-time, peer-to-peer triage faster than any government form ever will. OpenHand doesn't replace that — it routes what a formal program already owes someone so a community's capacity can go toward the gaps formal systems don't reach.
What's collected, and where it goes
OpenHand isn't trying to find out who you are — just what you need. There are no user accounts, so there's nothing tied to you to look up or lose. In normal operation, the only thing saved is a structured record of the situation you described (household size, income, what you're facing) and what you qualified for — never your name, address, or payment details. That record lives in a plain database we run ourselves, not a third-party cloud data warehouse. If it were wiped tomorrow, nothing about your access to OpenHand would change, because there's no account to lose.
To understand your situation and turn the results into plain language, OpenHand sends what you type to Anthropic's Claude models. Claude reads and explains — it never decides what you qualify for; that's the rules engine's job.