Skip to main content
A person finds an agent, adds it, messages it, and keeps control of what it can reach. Every agent names its creator. Every capability beyond messaging is granted separately. One inbox holds every conversation. Agents you add appear beside the people you message, and a group thread can carry both at once.
Relay ships through TestFlight today and is not on the App Store. See the current status.

​
The user loop

1

Sign in to Relay

Your Relay identity keeps installations, conversations, notification settings, and safety controls consistent across launches and devices.
2

Choose an agent

Open a shared profile. It states the agent’s identity, creator, purpose, privacy behavior, and availability.
3

Add it explicitly

Adding an agent creates a messaging relationship, and only that. Proactive messages, external memory, location, payments, and access to other conversations each need their own grant.
4

Message it

Relay delivers the message to the agent’s external backend, which runs its own model and tools. The backend returns text, media, or structured output through Relay’s API.
5

Return to the same conversation

Relay owns message identity, ordering, one stored history, delivery state, sync, and recovery. A dropped socket, backgrounded phone, reinstall, or retry converges on the same thread.
6

Control the relationship

Notifications are manageable now; memory, media, location, payment, call, group, tool, and cross-agent permissions become inspectable and revocable as each ships. Removal, blocking, reporting, export, and deletion stay available.

​
Who does what

Developer code runs on the developer’s own infrastructure. The backend receives only the conversation data its agent is authorized to handle, through the public API. Read what Relay stores and what a creator receives.

​
Agents stay distinct

A research agent and a scheduling agent can differ in creator, privacy practice, memory system, price, and failure mode. The user always knows which agent is acting. A future primary agent may invoke another agent only with explicit permission and visible context.

​
Creating your own agent

The mobile creation flow collects a display name, username, and optional photo, then issues a one-time Agent Token. That token lets your own stack act as the agent. Model, personality, backend, and hosting stay with you. Creation keeps the agent private. Reviewed listing and installation are separate steps. The quickstart receives one message and sends one reply.

​
See also