Book the world.
Get paid for it.

Booking a trip is a tab-hoarding chore spread across a dozen sites, each with its own account, its own currency and its own loyalty scheme. Three things are broken at once, and they are related.

Search is fragmented. Flights on one site, the sleeper train on another, the match ticket on a third that only sells in the home currency. No single surface understands “get me there, and into the stadium”.

Loyalty is captured. Points are locked to one airline or one hotel group, they expire, and they are worth nothing anywhere else. The value you create by booking never becomes yours to move.

Money is opaque. You pay in your card’s currency, the merchant charges in theirs, and the spread disappears into a statement line you cannot audit.

One idea answers all three: an agent that searches everywhere, settles in a dollar-pegged unit, and pays you a portable token for every booking it handles.

BookGent AI is a front desk of 9 agents, each a specialist in one kind of travel. You pick a desk, describe the trip, and the agent does the finding. It is not a metasearch frame and not a chatbot bolted to a booking API.

Nine desks and not one, because the question a ferry crossing asks is nothing like the question a matchday asks. Each desk owns its own search menu, its own vocabulary and its own reward rate, from 3.2% to 8.5% of the booking. A crossing and a cup tie are not worth the same to arrange.

Every result an agent returns carries its source, and the two sources are never blended, because they mean different things about who touched the money.

Path A hunts the open web. The agent runs a live search, reads the pages it finds, extracts real fares and converts each into USDG at a live rate. Each result is an outbound link to the merchant’s own site, where you complete the booking. The money is theirs. No $Bookai is paid, because paying a reward on a sale BookGent AI did not handle would be a lie.

Path B settles here. When the chain integration is live you book inside BookGent AI: a contract settles the fare in USDG on Robinhood Chain and the reward engine pays $Bookai in the same motion. This is the path that pays, because this is the path where BookGent AI is the merchant of record.

A fare on someone else’s site is theirs to sell. The two paths stay visibly apart so a reward is never claimed for a sale this platform did not settle. That principle is not negotiable, and it is the reason one of the two paths pays nothing.

BookGent AI settles on Robinhood Chain, an Arbitrum Orbit L2 with gas in ETH, and prices everything in USDG, the Global Dollar issued by Paxos and pegged 1:1 to the US dollar.

USDG is the stablecoin native to this chain, so here it simply is the dollar. Pricing fares in it means the number a traveller reads and the number that settles are the same number, with no rate to apply afterwards and no spread to hide inside. Converting a €200 fare means applying a live rate, not relabelling the number. The merchant’s own price sits beside the result, so the arithmetic can be checked instead of taken on trust, and a fare that cannot be converted is dropped rather than guessed at.

Network

Robinhood Chain · Arbitrum Orbit L2

Mainnet chain ID

4663

Testnet chain ID

46630

Gas

ETH

Settlement unit

USDG, the Global Dollar from Paxos

Explorer

Blockscout

$Bookai turns the value a traveller creates by booking into something portable they own. It is earned by booking, never sold to unlock a feature.

base        = fare × the desk's own rate   (3.2% – 8.5%)
tier bonus  = base × (your multiplier − 1)   (up to ×1.6)
crew bonus  = 15 $Bookai   (first booking with a new desk)
─────────────────────────────────────────────
reward      = base + tier bonus + crew bonus

Your tier is set by lifetime $Bookai earned, and it scales every future payout, so the reward rises with how much of the world you have booked, not with how much you hold.

The first booking you make with a desk you have never used pays a one-off 15 $Bookai finder’s bonus. Working all 9 desks is the long game the reward system is built around: it pays for breadth of travel, not for spend.

The mechanics below are fixed by pons and are the same for every token it deploys. They are stated here because they are already true.

Total supply

1,000,000,000, minted once and never again

Decimals

18

At creation

The whole supply went to the bonding curve. No allocation held back, by anyone.

Paired with

ETH

Creator tax

3%, fixed at creation, on top of the protocol trading fee

Graduation

4.2 ETH into the curve, then a locked Uniswap v4 pool

$Bookai launched on 31 August 2026, on pons, operated by Pons Labs, LLC. BookGent AI has no partnership with pons and claims none: it launched there the same way anyone else can.

pons is non-custodial: your own wallet submits every transaction, which is how this application already works. v2 launches on a bonding curve holding the whole supply, and the pool is built only once that curve sells out.

The opening moments are priced, not gated. A buy in the first instant pays away 99% of itself, and the charge decays to nothing across 5 seconds: roughly 25% one second in, about 3% at two. It falls on buys alone, selling is never taxed, and what a racing buyer gives up is shared out like any other trading fee, so it returns to the launch instead of leaving it.

pons states plainly that graduating is not a quality signal and guarantees no future liquidity, price, or exit. Repeating that here is more use to a reader than leaving it out.

BookGent AI is a Next.js application with a clean seam between the interface and every external system, so each dependency can be swapped without touching anything above it.

The hunt runs behind a single provider seam: a language model writes the queries, a search layer fetches the pages, and the extraction reads them. Swapping in a model that browses natively changes nothing the interface can see. Failures are loud. A rejected API key produces a user-facing error rather than a quiet fallback to something invented.

The wallet is written against an adapter interface with EVM-shaped types, so moving from this platform’s own ledger to a live adapter pointed at chain 4663 needs no change to the reward engine above it. Connection is real today: the site asks your own browser wallet for an address and never holds keys.

Reference data binds every operator to the geography it truly serves, and distances are measured between real coordinates, not region centres. Real names appear only on real routes: using an operator’s name on a route it does not run is worse than an invented name, because it looks authoritative while being false.

The honest status, in the place a reader is most likely to assume otherwise.

• Wallet connectionliveReal. Your own browser wallet, asked for an address and nothing more.
• The desks and their menusliveAll 9, with reference data bound to real geography.
• Live web searchlivePath A returns real fares from operators' own pages, converted and linked out.
• On-chain settlementnot yetFares do not yet settle in USDG on Robinhood Chain.
• $Bookai on chainnot yetThe contract is live. This wallet figure is still this platform's own ledger, not a chain balance, until settlement is wired to chain 4663.
• Desk faresnot yetPrices on the desk path are illustrative until desk partners go live.

Nothing in this document is an offer to sell a token, a distribution commitment, a price expectation, or investment, legal or tax advice.