🤫 Private Agent One — mission and purpose

TL;DR: Private Agent One exists so Americans can seamlessly connect, communicate, and conduct commerce, charity and well-being work across their communities — through agents who are humans, working at brands, businesses, charities, and the institutions that help people pay their bills.

Status as of 2026-07-28: see body.

The mission

🤫 Private Agent One is the American citizen's own agent — sovereign over their own data, agents and compute — and its job is the day-to-day private and personal agent work they would love to hand to somebody else.

The life admin. The chasing, comparing, forms, scheduling, renewals, follow-ups. The work that is nobody's idea of a life.

And the reason it matters is what the person actually wants: somebody sitting in between them and everyone else. Between them and all the human beings who serve them — their agent, their adviser, their doctor, their contractor, their caseworker. And between them and all the AI agents, apps and services they pay for and get genuine benefit, happiness and enjoyment from.

Serving that purpose is how Americans come to connect, communicate, and conduct extensive commerce, extensive charity, and extensive well-being work across their communities — leveraging agents who are humans, working at brands, businesses, charities, and the institutions that help people pay their bills.

Interposed, not a destination

This is the architectural claim, not a positioning line. Agent One is delegated to. It negotiates and transacts on the person's behalf under their consent, absorbs the interface burden, and shields their data and attention from everything on the other side of the line.

The design consequence is strict: every integration is something Agent One does for the person from their side, never another surface the person has to go and operate. A feature that makes the human do more work rather than less is the wrong feature, however impressive it is.

What is easy to miss in it

Commerce is one third. Charity and well-being work sit beside it as equals, not as a corporate-social-responsibility footnote appended to a sales motion. A network that only connects buyers to sellers has built a third of this.

The agents are humans. The AI agent serves the person; the Network is made of people — the insurance producer, the benefits caseworker, the volunteer coordinator, the credit-union officer. Private Agent One is how a person reaches them, not a replacement for them.

"Institutions helping them pay their bills" is a category, not a flourish. Banks, credit unions, utilities, benefits and assistance offices, financial-counselling nonprofits. For a great many households this is the most consequential connection in the list, and it is the one no directory has ever taken seriously.

Seamlessly is the standard. Fewest clicks to done. If reaching the right human in your own neighbourhood takes a search, three phone trees and a form, the mission has not been met however complete the data is.

How it serves a person: start at privacy, then work outward

Agent One serves someone by genuinely understanding them — their needs, wants and desires — and it starts at privacy preferences, because that is the one preference every website, app and brand on earth already legally needs from them. It is the wedge with a door already open.

From there it works outward into the ordinary texture of a life: food, clothing and how they dress, eating, brushing, their eyes, what they smell like and the perfume they wear. The cars they like, the music they like, the sports and the teams and characters they follow, the games they love to play. The work they do and the work they like to do. The thinking and the reading they choose. Everything that makes them specifically them rather than a demographic.

The shape of this is an inversion. Today a thousand brands each assemble their own shadow profile of the same human, none of it visible to them and none of it theirs. Here the person publishes one canonical, self-owned profile and the world subscribes to it — with granular read access they grant per field and per subscriber, revoke at any time, and see a receipt for on every access.

Two rules follow from that and they are not negotiable: never infer what the person has not told us, and never let a subscriber read a field they were not granted. A profile that guesses is just a shadow profile with better manners.

What it implies for the work

Any surface, dataset, or go-to-market motion that reaches only commercial businesses is missing two thirds of the mission. The directory and the underlying data model have to reach charities, community institutions, and bill-paying institutions with the same seriousness applied to licensed professionals.

Measured against that standard today, coverage of licensed commercial and clinical professionals is strong and coverage of charities, community institutions, and assistance organisations is effectively absent. That is a named gap, not an oversight — and the honest way to hold it is on the same public coverage page that reports everything else.

It runs 24/7/366, on what the person already owns

Always on, for one user, ideally on their own computer and their own infrastructure, using their own existing subscriptions they are already paying for.

This is why three things are native and first-class in the app rather than integrations bolted on later:

• Bring your own subscription (BYOS) — they already pay for it; we do not resell it back to them.
• Bring your own computer (BYOC) — their Mac, their phone, their 🤫 Puppy One, their garage grid.
• Bring your own API (BYOA) — their keys, their quota, their account.

Because the work runs on hardware the person owns with credentials the person holds, the data does not have to leave and the bill does not have to double. That is the technical spine under Own your AI. Own your data. Own your compute.

Google-first, Gemini-first

We are a Google-first, Gemini-first company. Google and Gemini are supported out of the box in the day-zero experience — not as one option among many a user has to go and configure. Day zero has to work for someone who arrives with a Google account and nothing else.

Open source and open models, wholeheartedly

We support every open model that meets or exceeds the standards the open-source community itself sets for standards, safety and trust.

Both halves of that matter. Every model that clears the bar — we do not pick winners, gate on commercial relationships, or quietly deprioritise a model because it competes with a partner. And the bar is the community's own: published model cards, disclosed training and evaluation practice, real licence terms, a maintainer who responds.

This follows from everything above rather than sitting beside it. An agent running on the person's own compute should be able to run a model they can inspect, host, fork and keep. A model the user can own is the only kind that fully honours Own your AI.

There is no tension with Google-first: the default has to work before a user has done anything, and the open ecosystem is what they graduate into when they want the whole stack in their own hands.

Free to every American citizen

🤫 Private Agent One is given, not sold, to the people it serves. Free for every American.

This is not a launch promotion or a freemium entry point, and it decides more than a price. If the revenue never comes from the citizen, it cannot come from their data, their attention, or their contacts either — those are the only other ways a free consumer product has ever been paid for, and each of them is the business model this company exists to refuse.

It is also what keeps the interposition honest. An agent standing between a person and everyone who wants to reach them becomes a toll booth the moment either side pays it.

What we sell is to businesses and institutions: compute they own, the agent platform, the robots. Never the person.

Concretely, any pricing page, paywall, premium tier, trial expiry, or upsell aimed at a citizen contradicts the mission. If a surface asks an American to pay for Agent One, it is wrong.

What stays true regardless

Consent first. Every listing claimable, correctable, and removable in one step, free, with no retention flow. A network someone was enrolled in without asking is not a network — and for a charity or a benefits office, being listed wrongly carries a cost a business would never feel.

Sources

• Founder statement of mission, 2026-07-28
• Directory coverage, measured and published