# Identity providers & sign-in

> How a human signs in to the platform — a local admin password on Desktop, or a connected OIDC provider on a cluster — and becomes the subject every request is authorized against.

Source: https://openap.org/docs/idp-signin

This is about *your* identity, not the agent's. Before you can open the dashboard or a chat, the
platform has to know who you are — and that signed-in identity becomes the **subject** every
request is authorized against in the same graph the agents use.

## Two ways in

- **A local admin password.** On [Desktop](https://openap.org/docs/install-desktop) — and any single-operator install — you set an
  admin password at setup and sign in with it. Simple, local, no external dependency.
- **A connected identity provider.** On a shared cluster you connect an **OIDC** provider (your company SSO) via
  a **`ClusterIdentityProvider`**, and people sign in through it. The install offers to set this up, or you
  configure it later with `oap idp`.

Either way, an unauthenticated request to a gated surface is redirected to sign in; only after that does the
request carry a real identity.

## From sign-in to a subject

A successful sign-in mints a session for the browser, and the **verified identity** — an email from the trusted
provider — is what becomes the SpiceDB subject on every subsequent request. So "can this person see this
session / install this agent / approve this step" is answered by the *same* per-subject check as everything
else. There's no second, weaker access model for the human side.

> **Security: The identity is verified, then checked — not asserted**
>
> The platform trusts an identity because a provider it configured vouched for it, and then still authorizes
> each request against the graph. A signed-in user isn't implicitly an admin; being able to do a thing is a
> permission someone granted, checked live — the same fail-closed discipline as agent authorization.
