OpenFirma x PAi
OpenFirmaPAi

Autonomy without blind trust.

PAi is useful because it can act. OpenFirma makes that safe: you set limits once, PAi stays inside them, asks when something crosses a line, and every action is readable and reversible where possible.

The tension

Every agent has the same problem.

Let it do anything

Maximum convenience, but you are blindly trusting software with money, identity, personal data, and actions that may be hard to undo.

Approve every step

Maximum safety, but now the assistant interrupts you constantly and you are back to doing the work yourself.

OpenFirma resolves the tradeoff: PAi gets room to act, but only inside rules you understand and control.

How it feels

Set personal-assistant rules, not technical policies.

You never write code, touch a config file, or see a policy language. The web app gives ordinary controls: sliders, toggles, dropdowns, allowed contacts, blocked services, and spending limits.

Spending

"Up to $150 per order, $500 per month on travel, and never start a subscription without asking."

Personal data

"Share my delivery address with booking sites, never share my ID document, and use my card only with known billers."

Messages

"Email known contacts directly, draft first for anyone new, and never accept terms for me."

Services

"Shopping sites can be active, banking is read-only, and anything new needs my approval."

Three outcomes

Every governed action becomes allow, ask, or block.

Inside your rules

PAi handles it quietly and tells you when it is done. Example: a $120 flight inside your travel budget.

Crosses a line

You get one clear request showing service, amount, recipient, and data shared. One tap approves just that action.

Against your rules

It does not run. PAi tells you why, and the assistant cannot talk its way past the limit.

Human in the loop

When something crosses a line, PAi asks you first, right in the chat.

PAi reads, drafts, and searches on its own. But before it spends money, sends something out, or shares your identity, OpenFirma pauses it and sends you a request in Telegram. You approve or decline with a tap. Nothing happens until you do.

No dashboards, no jargon. Just a clear request with the amount, recipient, and what is being shared, plus two buttons.

9:41
PAi
PAipersonal assistant
Today
Renew Acme Hosting annual plan.
9:40✓✓
Payment approval
Acme Hosting wants to charge your annual renewal. Approve the payment?
Amount€240.00
ToAcme Hosting
Card•• 4421
9:41
Message
9:42
PAi
PAipersonal assistant
Today
Draft Q3 follow up to John. Stop before sending.
9:41✓✓
Send email
I drafted your reply. Send this email to john@client.com?
Tojohn@client.com
SubjectQ3 proposal follow up
9:42
Message
9:43
PAi
PAipersonal assistant
Today
Get the TicketOne tickets. Stop before sharing ID.
9:42✓✓
Identity check
TicketOne needs to verify you before checkout. Share your ID document?
ShareID document
WithTicketOne
9:43
Message

Why PAi + OpenFirma matters

PAi makes OpenFirma tangible. OpenFirma makes PAi distributable.

PAi provides the personal runtime, identity, memory, browser sessions, and operational work. OpenFirma provides scoped authority, user-set limits, approvals, and audit. Together they create an assistant that can be genuinely useful without asking users to blindly trust it.

Control without babysitting

You do not approve every click. You set the bounds once and intervene only when the task needs judgment.

Safety without jargon

Capability tokens, policy engines, and audit ledgers stay behind the curtain. The user sees limits, requests, and history.

Trust that scales

An assistant you can trust with money, identity, and real-world actions is one you can actually hand work to.

Early access

Your assistant is ready when you are.

PAi is opening to a small group of early users. Request access and we will set yours up for you.