# Protect an Express API

This quickstart protects an Express 5 API with KeenAuth access tokens. You add the `expressAuth` middleware of the Node.js SDK, require a scope on one route and test the API offline with the SDK test kit. Express is not a dependency of the SDK; the middleware also works with Express 4.

The API has three endpoints:

| Endpoint | Requires |
|---|---|
| `GET /api/public` | nothing |
| `GET /api/private` | a valid access token for your API |
| `GET /api/messages` | a valid access token with the `messages:read` scope |

## Before you begin

- A KeenAuth environment. Its issuer looks like `https://your-environment.eu.keenauth.app`.
- An API registered in the environment ([Create an API](/docs/api/management/apis/create)). Its `identifier`, for example `https://api.example.com`, is the audience of the tokens your API accepts.
- The `messages:read` scope on that API, granted to the application that calls it.
- Node.js 22 or later.

## 1. Install the SDK

```sh
npm install @keenauth/node express@5.2.1
```

The SDK is not published yet. Until it is, the example links to the SDK source in its `package.json`.

## 2. Configure

The API reads its configuration from environment variables only. Put them in a `.env` file, which `npm start` loads with `node --env-file-if-exists=.env`, and never commit that file.

| Variable | Required | Meaning |
|---|---|---|
| `KEENAUTH_ISSUER` | yes | Your environment's issuer |
| `KEENAUTH_AUDIENCE` | yes | Your API's identifier |
| `PUBLIC_BASE_URL` | yes | The URL clients call (see [Behind a proxy](#behind-a-proxy)) |
| `KEENAUTH_WEBHOOK_SECRET` | recommended | The `idp_whsec_` signing secret of the webhook endpoint that delivers `token.revoked` (two, comma separated, while you rotate it). Without it the API never learns of revoked tokens |
| `PORT` | no | Listen port, default `8080` |

## 3. Protect your routes

`createApp` creates the protector and the routes. `expressAuth(protector)` requires a valid token; with `{ scopes }` it also requires scopes. The handler reads the verified claims from `req.auth`.

```js
export function createApp({ verifier, publicBaseUrl, revocation, webhookSecrets = [] }) {
  const protector = createProtector(verifier, {
    publicBaseUrl,
    revocation,
    // The error never contains the token or claim values; never log the Authorization or DPoP header.
    onError: (err) => console.warn(`rejected: ${err.message}`),
  });

  const app = express();
  app.disable("x-powered-by");

  // No token needed.
  app.get("/api/public", (req, res) => {
    res.json({ message: "Hello from a public endpoint." });
  });

  // Any valid access token for this API.
  app.get("/api/private", expressAuth(protector), (req, res) => {
    res.json({ message: "Hello from a private endpoint.", sub: req.auth.subject });
  });

  // A valid access token that also carries the messages:read scope.
  app.get("/api/messages", expressAuth(protector, { scopes: ["messages:read"] }), (req, res) => {
    res.json({ messages: ["Hello from a scoped endpoint."] });
  });

  // token.revoked deliveries from KeenAuth; the handler reads the raw body itself, so no JSON parser runs before it.
  if (webhookSecrets.length > 0) {
    app.post(WEBHOOK_PATH, ...createWebhookHandler({ secrets: webhookSecrets, revocation }));
  } else {
    console.warn("KEENAUTH_WEBHOOK_SECRET is not set: token.revoked events are not received, so revoked tokens stay valid until they expire");
  }

  return app;
}
```

`server.js` builds the real verifier from the configuration. The verifier fetches the signing keys from your environment the first time it sees a token.

```js
const issuer = process.env.KEENAUTH_ISSUER;
const audience = process.env.KEENAUTH_AUDIENCE;
// The URL clients call: DPoP proofs are checked against it, never against the Host header.
const publicBaseUrl = process.env.PUBLIC_BASE_URL;
if (!issuer || !audience || !publicBaseUrl) {
  console.error("KEENAUTH_ISSUER, KEENAUTH_AUDIENCE and PUBLIC_BASE_URL are required");
  process.exit(1);
}

// createVerifier refuses an issuer under the reserved .invalid domain, so a test token can never pass here.
const verifier = createVerifier({ issuer, audience });
const revocation = createMemoryRevocationCache();
// The signing secret (idp_whsec_...) of the webhook endpoint that delivers token.revoked; comma separated while rotating.
const webhookSecrets = (process.env.KEENAUTH_WEBHOOK_SECRET ?? "").split(",").map((s) => s.trim()).filter(Boolean);
const app = createApp({ verifier, publicBaseUrl, revocation, webhookSecrets });

const port = Number(process.env.PORT || 8080);
app.listen(port, () => console.log(`listening on port ${port}`));
```

## 4. Run and call the API

```sh
npm start
```

For a first call, get a token with the client credentials grant of a machine-to-machine application that has the `messages:read` scope on your API:

```sh
ACCESS_TOKEN=$(curl -s -u "$CLIENT_ID:$CLIENT_SECRET" \
  -d grant_type=client_credentials \
  -d resource=https://api.example.com \
  -d scope=messages:read \
  https://your-environment.eu.keenauth.app/oauth2/token | jq -r .access_token)

curl -i http://localhost:8080/api/messages -H "Authorization: Bearer $ACCESS_TOKEN"
```

## 5. Test without a network

`@keenauth/node/testing` mints tokens signed with a public test key, and `createTestVerifier` accepts only those tokens. The tests pass that verifier to `createApp` and call the app on a local port, so they need no network and no KeenAuth account. See [Offline test kit](/docs/sdk/testing).

```js
test("scope check: 403 insufficient_scope without messages:read, 200 with it", async () => {
  await expectProblem(await get("/api/messages", bearer(mintAccessToken(AUDIENCE))), 403, "insufficient_scope", /scope="messages:read"/);
  const ok = await get("/api/messages", bearer(mintAccessToken(AUDIENCE, { claims: { scope: "profile messages:read" } })));
  assert.equal(ok.status, 200);
});
```

```sh
npm test
```

The example's tests check that a valid token is accepted and that a missing, expired or wrong-audience token, a missing scope, a token in the URL query and a revoked token are rejected with the statuses below. They also send a DPoP-bound token with a proof for `PUBLIC_BASE_URL`, and check that `server. The tests also send signed `token.revoked` deliveries: a valid one revokes the token, a forged, tampered or stale one gets `400` and revokes nothing, and a repeated one has one effect.js` refuses to start with the test issuer.

## Responses

| Request | Status | `code` in the body |
|---|---|---|
| No `Authorization` header | `401` | `missing_token` |
| Expired token, token for another API, bad signature, revoked token | `401` | `invalid_token` |
| Token without the required scope | `403` | `insufficient_scope` |
| Token in the URL query (`?access_token=`) | `400` | `invalid_request` |

The body is `application/problem+json` and the `WWW-Authenticate` header carries the standard challenge. The full table is in [Protecting your API with middleware](/docs/sdk/protect-api#responses).

## Behind a proxy

If clients use DPoP, the middleware checks that each proof was made for the URL the client called. Set `PUBLIC_BASE_URL` to the external URL clients call, for example `https://api.example.com`, also behind a proxy or load balancer. It is required and must use `https` (`http` only for `localhost`, `127.0.0.1` or `[::1]`); the app refuses to start without it. The middleware never trusts `X-Forwarded-*` headers, whatever Express's `trust proxy` setting is.

## Revoked tokens

An access token stays valid until it expires, so KeenAuth tells your API when tokens are revoked. The example passes an in-memory revocation cache to the protector and serves a `token.revoked` webhook at `POST /webhooks/keenauth` (`webhook.js`) that feeds it.

1. In the dashboard, create a webhook endpoint with the URL `https://<your host>/webhooks/keenauth` that subscribes to `token.revoked`, and export its signing secret as `KEENAUTH_WEBHOOK_SECRET`.
2. The handler verifies the signature with the SDK on the raw body and answers `400` when it is invalid or the timestamp is more than five minutes off. It maps the event's selector (`access_token`, `session`, `user`, `client`, `org`, `org_user`, `client_user`) to a revocation marker and answers `204`; the middleware then rejects every matching token issued before `not_before`.
3. KeenAuth delivers at least once, so the handler applies each `webhook-id` once. If the marker cannot be stored it answers `503` and KeenAuth delivers the event again. It ignores other event types and selectors that no access token can match (`refresh_family`, `api`).

The handler logs the event id, the selector type and the reason, never the body, a signature or the secret. Each process has its own cache; with several instances, put the markers in a shared store behind the same `revoke` method. See [Revocation cache](/docs/sdk/protect-api#revocation-cache-tokenrevoked).

## Next steps

- [Protecting your API with middleware](/docs/sdk/protect-api): Fastify, step-up requirements, DPoP and every option.
- [Verifying tokens](/docs/sdk/token-verification): the verifier under the middleware.
- [Offline test kit](/docs/sdk/testing).