How do I give my agent an identity?

giving your agent an identity lets it act on your behalf. it can sign up for a service, accept its terms, receive its email, and use the account. the right way to do that is through an agent identity broker. with Agent Registrar, you verify your email address once, sign one delegation agreement, and give each agent its own identity under your account. when the agent reaches a service, the broker vouches for it. the service can check that the agent acts for you, and let it in.

the protocol behind Agent Registrar is open, so other brokers can do the same.

how not to give your agent an identity

most shortcuts look like delegation. they hand over more than the task needs, and they leave no record of what happened.

• put your name, address, phone number, and account details in a text file and paste it into the prompt. the agent can repeat any of it to any service that asks. you cannot see which service received which field, and you cannot take it back once it is in a log, a transcript, or another agent's context.
• give it your password manager, your browser profile, or a session cookie. that is not an identity. it is every login you have, with no scope and no per-service record. an agent that is tricked hands the attacker the same access.
• let it invent a persona with a throwaway inbox and a made-up name. the service gets a reachable address, not an accountable human. when something goes wrong, there is no one to contact and no accepted agreement to point to.
• use one API key or one shared login for every agent. the key may get them in, but it does not say which agent did what. revoking one agent means breaking the others.

what an agent can do with your identity

the first job is signup, because that is where most agents stop. when a service supports agent signups, it asks the broker to prove that the agent acts for a verified human. the broker provides that proof, the service creates the account, and the agent can finish the signup without you filling in the form.

the terms agreement is the other common blocker. many services will not let an agent in until a human accepts their terms of service. with a broker, your agent can accept those terms under the delegation you signed. the broker keeps a record of that acceptance, and the service can check that you authorized the agent to accept on your behalf. then the signup can continue.

the agent also gets its own inbox, so it can receive the verification email and service notices. you do not have to forward codes by hand. after signup, it can use the account with the credentials the service gives it. each agent has its own identity, so you can keep them separate and revoke one without touching the others.

payments need an accountable owner

spending money raises the same question in a sharper form. a payment method tells a service how the agent can pay. it does not tell the service which human is accountable for the spend. the Machine Payments Protocol (MPP) is an early standard for agents paying for things online. it can check which software made a payment request. it cannot check which human is behind the agent or what the agent is allowed to buy.

the identity question is the same at payment time. the broker's job is to make the human behind the agent checkable. the payment method stays a separate choice, and Agent Registrar does not move money or hold a payment method.

why use an agent identity broker

without a broker, the exchange is ad hoc. the service asks for whatever it wants, and the agent supplies whatever it has. nothing in that exchange says which human is behind the agent, which terms were accepted, or how to stop the agent later.

a broker gives the exchange a shape. the service gets proof that the agent acts for a verified human, not a copy of your personal file or your password. the service can check that proof. you can check the record.

that shape is the agreement about what the service asks for and what the agent shares. the service decides which brokers it will accept. you decide which agents to create, what they can do, and when to revoke them. the service still owns the account and the credentials. you still own the decision to delegate.

registration is what makes the delegation durable. a registered agent has its own identity and inbox. that is what lets you give it narrow access and revoke it later. without that, a delegation is just a sentence in a prompt.

the broker verifies your email address, not your legal identity. the accountability comes from the delegation and the record.

the broker handles the identity and the record. the service keeps the account and the credentials. a broker does not make a service trustworthy. a service can still misuse the contact details it receives. the trust and limits page covers that.

to give an agent an identity under your account, start with the agent quickstart. it walks through connecting a harness, registering the agent, and completing a first signup.