For human app users
Give your AI agent its own limited access to the apps you use. You choose what it can do and for how long, and you can remove it with one tap. Your password never leaves your hands.
The problem
AI agents can now book, buy and check accounts for you. But apps only have a front door built for people, so your agent has to borrow your keys.
You hand over your password or let the agent drive your browser. Then it can do everything you can, with no limits.
There’s no list of what it looked at, changed or bought, and no way to take access back short of changing your password.
Courts are still sorting it out. In March 2026 Amazon won an order blocking Perplexity’s shopping agent. In August a federal appeals court threw it out, treating the person using the agent as the one responsible. If it’s on you, you should control what it can do. Cooley
How it works
It all happens inside an app you’re already signed in to, like your health plan, bank or utility. No new account, and no password to hand over. You tap “Invite your agent,” and IDhen handles the rest.
It sits inside the apps you already use, next to tasks like refilling a prescription.
Choose the agent you already use. It only gets what you approve next.
Face ID, a passkey or a one-time code. Works on any device.
ChatGPT confirms it’s your account on its side.
Northstar sets the limits. Within them, you pick what’s allowed (view claims, schedule refills, message your care team), and some things stay off limits (change your address, see payment info). Set a time limit, and see that the agent is verified, and whose it is, before you approve.
It can check your claims, schedule refills and message your care team, within the limits you set. No password changed hands.
Your rules
Trust & security
Only what you grant, for as long as you allow. Sensitive actions can be blocked.
See which agent is asking, and who owns it, before you approve.
Every permission you give is saved, with when you gave it.
This session, 24 hours, 30 or 60 days, or ongoing. Change it or end it anytime.
Remove one agent, or every agent everywhere, in one tap.
Built on open security standards, not a closed system.
FAQ
A safe way to let your AI agent help you inside the apps you already use. It gets limited, time-boxed access, and you can change or remove it any time. Look for the “Invite your agent” button.
No. It gets its own limited seat in the app, with only the permissions you approve.
Only what you approve, within limits the app sets. Some actions, like changing your address, are never open to agents.
It doesn’t get through. Its access only covers what you approved, and the app turns down anything else.
Remove its access from your IDhen dashboard or from the app. You can also remove every agent everywhere at once. If you picked a time limit, access also ends on its own when it runs out.
Before you approve, you see the agent’s name, who it belongs to and that it’s been verified. If anything looks off, tap Deny.
The goal is any personal AI agent that can prove who it is and who it works for. We’ll share which ones come first as we get closer.
In the kinds of accounts where people already get help from family or caregivers: health plans first, then benefits, utilities and phone accounts.
Not yet. Join the waitlist and we’ll email you when you can try it.
For app developers
Agents already come in through the side door, with shared passwords, forwarded codes and scrapers. You can’t tell them from your user, or from an attacker. IDhen lets agents in safely with one integration and ends the password sharing and scraping. You know which agent acts for whom, with proof of what the user allowed.
The problem
Personal AI agents now book, buy and check accounts for people. Most get in with a shared password, a forwarded code or a scraper driving a browser, so to your systems they look like the user or like an attacker.
Your users want their agent to refill the prescription or pull the statement. Shut it out and they find workarounds, like handing over their password.
Shared credentials mean no scopes, no expiry and no way to tell the agent from the person. When something goes wrong, nobody can show which agent had access or what it was allowed to do.
Courts are still sorting it out. In March 2026 Amazon won an order blocking Perplexity’s shopping agent. In August the Ninth Circuit threw it out, treating the person using the agent as the one responsible. Your app still needs to know who sent an agent and what it was allowed to do. Cooley
Some teams are already wiring up their own agent access, or thinking about it. A scope and a consent screen get one agent through. The full job is bigger:
IDhen is that layer, ready to drop in, so your team can stay on the product.
Already working on it? Talk to us about fitting IDhen around what you’ve built.
How it works
Put “Invite your agent” on any screen where users hand off a task, like a refill or a bill payment.
List what agents may do in your app and what’s never allowed. IDhen turns that into a plain-language consent screen and holds agents to it.
You set the longest access each permission allows. Agents can ask for less, and people choose within your limits. Users or your admins can revoke access at any time, and a webhook tells your app when they do.
<IDhenInvite
scopes={[
"claims.read",
"refills.schedule",
"careteam.message",
"records.download",
]}
maxDuration={{
"claims.read": "ongoing",
"refills.schedule": "30d",
"careteam.message": "60d",
"records.download": "session",
}}
blocked={[
"address.write",
"payment.read",
]}
onGrant={saveGrant}
onRevoke={endSession}
/>
// Lengths: "session", "24h", "30d",
// "60d" or "ongoing". An agent can ask
// for one (requestedDuration: "60d");
// IDhen caps it at each maxDuration.Not a live SDK. The API is being shaped with design partners.
Why it works
Sign in with Google and Apple taught people to log in through an account they already trust. PayPal did the same for payments, so a store never sees your card number. More recently, Plaid made linking an app to a bank account routine. Each time, one familiar consent step replaced a pile of one-off setups. IDhen brings the same model to agents.
One integration covers every agent, the way Plaid gave fintech teams one integration instead of one per bank. Scopes, expiry and revocation come standard, with no custom work per agent.
A step they already know. Pick the agent, see what it can do, approve. Nothing new to learn, and their password never leaves your app.
The user’s side
The whole flow runs inside your app, behind the sign-in your users already have.
The button sits on your own screens, next to the task it helps with. Users start inside your app, signed in the way they always are.
Only verified agents appear in the picker, so you know which agent is coming before it reaches you.
The yes comes from your real user, confirmed on their own device. No shared passwords, no forwarded codes.
The agent confirms the account on its side, so you know whose agent it is. One integration covers every agent, instead of one per agent.
Your rules shape this screen: which permissions exist, which stay blocked, and the longest each can last. Users choose within them, with your brand in front of them.
Your user’s agent can now handle real tasks in your app for them, like checking a claim or scheduling a refill. It works only within the scopes you offered, with a revocable token instead of a password.
What you get
One integration gives you consent, scoped permissions, verified agent identity, proof of what each user allowed and revocation.
Add the “Invite your agent” button and a ready consent screen. No auth or consent UI to build.
Each agent gets a token limited to the scopes you offer and the user approves, for no longer than you allow.
When a user or an admin removes access, the token stops working and your app gets a webhook.
Users can remove an agent at any time. Your admins can revoke one agent, one user or everyone.
Built on OAuth and the MCP authorization spec, alongside your existing login. It replaces nothing.
Know which agent is calling, which person it acts for and what that person allowed, on every request.
Designed for HIPAA-regulated workloads from the start. SOC 2 is on our roadmap.
Placement
Put “Invite your agent” where the chore already lives, not buried in settings. People think of handing a task off while they’re doing it.
These are starting points. Where Invite goes in your app is your call. You know your users and your screens best.
Next to a task people repeat, like the refill button or the bill pay form. They see it at the moment they’d hand the task off.
A “Connected agents” page in account settings, where people can invite an agent or remove one.
A light nudge once a task is done, such as “Next time, let your agent do this.”
Who it’s for
FAQ
A consent and identity layer for AI agents that act for your users. You add an “Invite your agent” button, users approve limited, time-boxed access, and each agent gets its own scoped token that can be revoked at any time.
No. IDhen sits on top of OAuth, the MCP authorization spec and the sign-in you already have. It adds the consent screen, verified agent identity and revoke controls that those standards leave to you. Okta MCP
You list the actions agents may take in your app and the ones that are never allowed. Users choose from your list and pick how long, within the longest time you allow for each one.
The agent’s token stops working right away, and your app gets a webhook so it can end any open session.
Not yet. IDhen is designed for HIPAA-regulated workloads from the start, and SOC 2 is on our roadmap. We’ll be specific about status with design partners.
Yes. Users can remove their own agents, and your admins can revoke one agent, one user’s agents or all agent access in your app.
Not yet. The code on this page is a sketch. Design partners help shape the API before it ships.
Working sessions with your product and security teams to map scopes, the consent flow and revocation for your app, then an early pilot when it’s ready.
WaitlistEarly access
Join the waitlist and we’ll email you when IDhen is ready to try. That’s the only email you’ll get from us.
We’re looking for a small group of product, platform and security teams, starting in healthcare and fintech. Help shape the scopes, consent flow and admin controls for your app, and get early access.
Thanks for your interest in IDhen. We’ll be in touch about early access.