Skip to main content

Protocol Core Concepts

This guide explains the protocol in product terms instead of endpoint terms.

If you only read one conceptual page before integrating, make it this one.

Concept 1: intent-first coordination

The center of OpenSocial is the intent.

An intent is a user’s coordination goal, such as:

  • meeting someone relevant
  • organizing a plan
  • starting a conversation
  • creating a recurring circle

Requests, chats, circles, and agent flows all exist to help that intent move forward.

Concept 2: app identity is not delegated authority

A registered app can authenticate successfully and still be unable to act for a user.

That is because the protocol separates:

  • app identity
  • delegated authority

Concept 3: actions are stable, not unlimited

The protocol exposes a small set of supported actions because those actions can be:

  • documented clearly
  • enforced consistently
  • observed in events
  • replayed or recovered operationally

That is better than exposing a wide, weak contract.

That also means some product workflows stay intentionally internal even when they are important to OpenSocial itself.

Today, that includes:

  • dating-consent workflows
  • commerce listing and offer workflows
  • profile editing and media-management flows
  • scheduled-task automation management

Concept 4: events are first-class

A protocol is not complete when an HTTP write succeeds.

It also needs:

  • delivery state
  • retries
  • dead letters
  • replay
  • cursors

This is why operations are part of the public docs.

Concept 5: agents consume the same protocol

The agent layer is not a separate protocol.

Agents use the same protocol surface as other integrations, with convenience wrappers for:

  • actor identity
  • readiness checks
  • tool catalogs
  • toolkit helpers

That keeps the architecture consistent and easier to trust.

The whole model in one diagram

Continue with