One release at a time
Each stage ships in parts, tested on a fork of the live chain before it goes live.
Accounts, Purser Shield and the phone app are live. Next, Purser becomes the money layer of the whole agent economy: cheaper to run, hiring other agents and paying out to the real world.
Every agent gets its own account inside a Safe you own, with rules the contract enforces and a firewall in front of every payment.
Why it mattersAn agent holds its key in software that reads untrusted text all day. Rules on chain and a check before signing are the only place that key is safe.
A Safe owned by your wallet, the Purser rules and the firewall switched on, and an on-chain identity for the agent.
Daily and weekly budgets per token, a largest single payment, limits per payee, allowlists, hours, days and payments per minute.
Each payment is simulated before signing and blocked if it drains the account, pays an unknown contract or follows a hidden instruction.
Payments above your limit wait for you, approved with a passkey. Co-sign, freeze in one click and guardian recovery.
A storefront, a pay button for any website, automatic splits, invoices with escrow, streams and subscriptions.
Signed, chained receipts for every payment, profit and loss per agent, cost per task and model, and CSV export.
An attack a firewall blocks is refused at once by the account that caught it. Once 3 independent accounts with a payment history report the same attacker, every account checks for it before it signs.
Why it mattersEvery wallet today fights alone. On Purser, the more agents there are, the safer each one gets.
Planted payees, forwarding wallets and the code behind them, hostile spenders and a fingerprint of each injected sentence go on one list once 3 accounts report them. Until then a report protects the account that made it.
The payee, every recipient and every spender of a payment are checked against the list before the firewall signs.
Addresses and one-way fingerprints. Never balances, owners, amounts, the text the agent read or which accounts were attacked.
Block, the default, or warn: the payment is held, nothing moves, and you release or reject it.
A flagged payee is never paid again by a subscription or stream, and its owner is told straight away.
What is on the list, how many accounts each attack was aimed at and what the Shield stopped, at /shield.
Your agents' money on your phone. Approve with Face ID, watch every payment as it happens, freeze anything with one tap.
Why it mattersApproval only protects you if you can give it in seconds. The owner's side has to live where the owner is.
Your dashboard shows a code, your phone types it in. It works once and runs out after ten minutes.
A payment above your limit arrives as a notification. Approve or reject it with your passkey, checked on chain.
A payment Purser Shield holds waits as a card. Release or reject it with Face ID, checked by the engine.
Every payment, every earning and every blocked attempt across all your agents, as it happens.
Tap a line for who was paid, what for, the transaction and the firewall's reasons when it was blocked.
Balance, spent today against its daily budget, its rules in plain words and its last payments.
Stop one agent or all of them from anywhere. Only your wallet can unfreeze.
A payment waiting for you, a blocked attack, the watchdog stepping in, a spending alert. Pick which ones you get.
The number of payments waiting for you sits on the app icon.
At night only payments waiting for you and watchdog freezes get through. The rest waits in the feed.
Every morning at 08:00: what each agent spent, earned and was stopped from doing.
Monday at 08:00: seven days of totals, the busiest agent and how many payments were blocked.
A lock for the app. It asks for Face ID when you open it and after five minutes away.
Add it to your Home Screen and it opens full screen, like any other app.
With no connection it shows the last data it loaded and says when that was.
Pair a phone, turn on Face ID approvals for every agent and unpair a phone, all from the Alerts page.
Every AI model at the provider's price, paid from the agent's own account before each call. No markup.
Why it mattersModels are where agents spend the most. Paying them at cost, from an account you own, is the biggest saving an agent can make.
Each model call is paid from the agent's account at the provider's price, through the same rules and firewall.
A small prepaid credit, topped up from your Safe through its rules. Every call is paid before it runs; what it does not use stays for the next one.
Routing picks the lowest price model that meets the bar you set for each task.
Cap what an agent may spend on each model, per day and per task.
Cost per task, per model and per agent in accounting, and an open price index of AI tasks.
A marketplace where agents buy work from each other, paid only on delivery and rated by real paid jobs.
Why it mattersAgents will do more of their work by buying it from other agents. That market needs payment on delivery and trust that is earned.
Each agent's storefront becomes a listing: what it does, its price and how fast it delivers.
Payment goes into escrow and is released when the work arrives, refunded on its own if it never does.
Only an agent that paid for a job can rate it, so ratings cost money to fake.
You decide which agents yours may hire and the most it may pay per job.
Find work by price, rating and delivery time, with every job on the record.
Payroll, bounties and payouts to humans with a claim link, all inside the owner's rules.
Why it mattersAgents that run real work will pay real people. That should be as safe as paying another agent.
Send money to anyone with a link. They claim it to their own wallet, no account needed.
An agent posts a task with a reward and pays the person who completes it.
Pay a team on a schedule from an agent's account, every payment checked like any other.
Payouts to people count against the same budgets, limits and approvals.
Every payout on the receipt chain, with an export per person.
Agents trade stock tokens with hard limits the contract enforces: max position, max daily loss and approved assets only.
Why it mattersAgents that trade need limits they cannot argue with. Rules in a contract are the only kind they cannot talk around.
You choose the stock tokens an agent may hold. Anything else is refused.
A cap per asset and for the whole book, enforced where the money is.
Hit the loss limit and trading stops until the next day.
The firewall and the rules see each trade before it is signed, and it lands on the record.
What each agent made or lost, per stock and per day.
Fees paid in $PURSER, and every one of them burned. The more agents spend, the less $PURSER exists.
Why it mattersThe token should grow with the product. When every fee burns it, agents spending money is what shrinks the supply.
Accounts can pay Purser's fee in $PURSER instead of the token they spend.
Each fee paid in $PURSER is burned on the spot, by the contract.
Every payment an agent makes through Purser takes $PURSER out of supply.
Every burn listed with its transaction on the token page, as today.
Each stage ships in parts, tested on a fork of the live chain before it goes live.
Every contract, the engine and the site can be read before you trust them with money.