Quick fix: n8n returns 404 with "The requested webhook ... is not registered" when nothing is listening on that exact method and path. If you called a /webhook-test/ URL, open the workflow and click Execute workflow (or Listen for test event on the Webhook node) first; the test URL then works for one call, for up to 120 seconds. If you called a /webhook/ URL, the workflow must be published (the Active toggle on n8n 1.x). Then send the request again with the method the node expects:

curl -i -X POST https://your-n8n.example.com/webhook/orders \
  -H "Content-Type: application/json" -d '{"id": 1}'

The exact error

n8n answers with HTTP 404 and a JSON body. These are the real responses from n8n 2.47.7 on our machine, with the debug-only stacktrace field removed. A production URL with nothing registered:

HTTP/1.1 404 Not Found
Content-Type: application/json; charset=utf-8

{"code":404,"message":"The requested webhook \"POST orders\" is not registered.","hint":"The workflow must be active for a production URL to run successfully. You can activate the workflow using the toggle in the top-right of the editor. Note that unlike test URL calls, production URL calls aren't shown on the canvas (only in the executions list)"}

A test URL with nothing listening (note that the method is missing from the message):

{"code":404,"message":"The requested webhook \"orders\" is not registered.","hint":"Click the 'Execute workflow' button on the canvas, then try again. (In test mode, the webhook only works for one call after you click this button)"}

The right path called with the wrong method (no hint this time):

{"code":404,"message":"This webhook is not registered for GET requests. Did you mean to make a POST request?"}

Where you see it: as the response body in whatever sent the request (curl, Postman, Stripe's or GitHub's webhook delivery log, an HTTP Request node), and in n8n's own log as Received request for unknown webhook: The requested webhook "POST orders" is not registered. In Docker, read that log with docker logs <container>.

Versions: the strings live in packages/cli/src/errors/response-errors/webhook-not-found.error.ts and are identical in n8n 1.123.96, 2.47.7 (current stable) and the 3.2.2 pre-release. Up to n8n 1.93.0 the test hint said "Click the 'Test workflow' button"; it became "Execute workflow" in 1.94.0 (May 2025). The stacktrace field only appears when NODE_ENV is unset or development; the official Docker image sets NODE_ENV=production, so Docker users see the shorter body.

Why it happens

Every Webhook node has two URLs. The test URL (/webhook-test/<path>) is registered only while the editor is waiting for a test event, and n8n's docs say it "stays active for 120 seconds". The production URL (/webhook/<path>) is registered when you publish the workflow and removed when you unpublish it. When a request arrives, n8n looks up the exact HTTP method plus path. If no published workflow (or no waiting test) owns that pair, you get the 404 above. If the path exists but only for other methods, you get the "Did you mean" version instead.

One terminology trap: since n8n 2.0 the editor has a Publish button, not an Active toggle, and the docs say production URLs work "until workflow is unpublished". The server hint still says "active" and "toggle". On 2.x, read it as "publish the workflow".

Find your cause

CauseHow to recognise itFix
Test URL, editor not listeningURL contains /webhook-test/; hint says "Click the 'Execute workflow' button"Click Execute workflow, then call within 120 seconds, once
Production URL, workflow not publishedURL contains /webhook/; hint says "The workflow must be active"Publish the workflow (n8n 2.x) or switch it to Active (1.x)
Wrong HTTP method"This webhook is not registered for GET requests. Did you mean to make a POST request?"Send the method the node expects, or allow several methods
Path typo, case or double slashThe path in the message is not the one on the node, for example "POST Orders" or "POST /orders"Copy the URL from the node again
Queue mode or several instancesWorks on one server or container, fails through the load balancerRoute /webhook-test/* to the main instance; share one database
Reverse proxy changes the pathPlain-text Cannot POST /n8n/webhook/orders instead of JSON, or the path in the message has an extra prefixForward the path unchanged, or serve n8n on its own subdomain

Cause 1: test URL without clicking Execute workflow

Open the workflow, click Execute workflow on the canvas (or open the Webhook node and click Listen for test event), then send your request. The docs list the test URL's listening duration as 120 seconds, and the hint itself warns that "the webhook only works for one call after you click this button". A second call, or a call two minutes later, returns the 404 again. For anything automatic (Stripe, GitHub, a form), use the production URL instead.

Cause 2: production URL, workflow not published

On n8n 2.x, click Publish in the editor; if you change the workflow later, publish again so the live version picks up the change. On 1.x, switch the Active toggle on. From the command line (self-hosted), n8n 2.x has publish:workflow:

n8n publish:workflow --id=<workflow-id>
# Docker:
docker exec -u node -it <n8n-container> n8n publish:workflow --id=<workflow-id>

We ran this against a test workflow: before publishing, POST /webhook/orders returned the 404; after publishing it returned {"message":"Workflow was started"}; after n8n unpublish:workflow the 404 came back within seconds. If publishing fails or shows "Partial publish", open the workflow: n8n only allows one webhook per path and method, and a clash stops the trigger from registering (see Related errors).

Cause 3: wrong HTTP method

The Webhook node accepts one method by default. Either send the one it shows in HTTP Method, or open the node's Settings, turn on Allow Multiple HTTP Methods, and add the methods you need (n8n's common-issues page describes these steps). Browsers send GET, so pasting a POST webhook into the address bar always triggers the "Did you mean to make a POST request?" message.

Cause 4: path typo, letter case or double slash

Matching is exact and case-sensitive. In our tests, /webhook/Orders and /webhook/order both failed against a node with path orders, and /webhook//orders failed as "POST /orders". A single trailing slash is fine: n8n strips it, and /webhook/orders/ worked. Copy the URL from the node's Production URL tab rather than retyping it. If you changed the path after publishing, publish again.

Cause 5: queue mode or more than one n8n instance

In queue mode with webhook processors, n8n's docs say /webhook/* can go to the webhook pool, but "the default URL for manual workflow executions is /webhook-test/*. Make sure that these URLs route to your main process." If your load balancer sends test calls to a webhook processor, they fail. Every main, worker and webhook process must also use the same database, so they all see the same published workflows. Two separate n8n containers with separate SQLite files behind one load balancer will answer 404 on whichever copy did not publish the workflow.

Cause 6: a reverse proxy rewrites the path

n8n serves webhooks at /webhook/ and /webhook-test/ from the root of its host. If your proxy forwards /n8n/webhook/orders without stripping the prefix, the request never reaches the webhook handler: our test returned an HTML page saying Cannot POST /n8n/webhook/orders (and a GET to the same path returned the editor page with status 200). If the proxy strips or adds segments, the path in the error message shows what n8n actually received. n8n's docs warn that combining N8N_PATH with reverse proxies "can cause folder navigation issues" and suggest a subdomain such as n8n.example.com. If the URLs in the editor show localhost:5678 instead of your domain, that is a different problem; see fixing n8n webhook URLs that show localhost behind Cloudflare.

Check it is fixed

Send the request with verbose output and the method from the node. On macOS and Linux:

curl -i -X POST https://your-n8n.example.com/webhook/orders \
  -H "Content-Type: application/json" -d '{"id": 1}'

In Windows PowerShell type curl.exe, because plain curl there is an alias for Invoke-WebRequest (the curl project documents this), and put the command on one line. With the default response mode you should see HTTP/1.1 200 OK and:

{"message":"Workflow was started"}

Production runs do not appear on the canvas. Check the workflow's Executions tab for the new run.

  • This webhook is not registered for GET requests. Did you mean to make a POST request? The path exists but not for that method (Cause 3).
  • The method PROPFIND is not supported. The method is not one n8n webhooks accept at all. In our test this came back as HTTP 500 with "code":0, not 404.
  • The URL path that the "Webhook" node uses is already taken. Please change it to something else. Shown when publishing: another published workflow owns the same path and method, so this one cannot register.
  • Received request for unknown webhook: ... The server-side log line for every 404 above; search your logs for it to see which paths callers are hitting.

What to do next

FAQ

Why does my n8n test webhook only work once?

By design. Clicking Execute workflow registers the test URL for a single call, and it expires after 120 seconds if nothing arrives. Click it again before each test call, or switch to the production URL.

Why does the production webhook return 404 when the workflow is published?

Check the method and the exact path first; the message shows what n8n received, for example "POST order". Then check that the request reaches the same n8n instance and database where you published it, and that no proxy changed the path.

Is "active" the same as "published" in n8n?

For webhooks, yes. n8n 1.x used an Active toggle; n8n 2.x replaced it with Publish and Unpublish. The 404 hint still uses the old word "active".

Can I use the test URL in production?

No. It only works while someone has the editor waiting for a test event. Give external services the production URL.

How this fix was verified

Run on macOS 26.5 (Apple M4) on 14 November 2026 with n8n 2.47.7 started through npx on Node.js 24.22.3 (a local npm package in a scratch folder, nothing installed globally), with N8N_USER_FOLDER in that folder, telemetry off and n8n bound to 127.0.0.1 only. We imported a one-node workflow (Webhook, POST, path orders) with n8n import:workflow, published and unpublished it with the CLI, and sent curl requests.

Reproduced here: every 404 body quoted above (unknown production path, unknown test path, wrong method, typo, capital letter, double slash, unpublished workflow), the 200 response after publishing, the trailing-slash behaviour, the Cannot POST page for a prefixed path, the PROPFIND 500, the Received request for unknown webhook log line, and the absence of stacktrace with NODE_ENV=production. We stopped n8n afterwards and confirmed port 5678 was free.

Checked against source and docs only: the wording in n8n 1.123.96, 1.93.0, 1.94.0 and 3.2.2 (GitHub source at those tags), the "already taken" message (n8n source), queue-mode routing and the 120-second test window (docs.n8n.io), and the clicks inside the editor UI, which we did not drive.

Last verified: 14 November 2026.

We re-check this fix monthly and record any change in this page's changelog.