How to Fix “The requested webhook is not registered” in n8n (404)
n8n gives you 2 webhook URLs that look almost identical. One stops working the moment you look away.
On this page
It worked when you tested it. The sender now gets a 404 every time:
404 - The requested webhook "POST my-hook" is not registered.
Nothing changed in the workflow. You can open n8n, click to listen, fire the same request, and watch it arrive perfectly. Then it stops again.
That pattern — works while you watch, fails when you don't — is the signature of this bug, and it narrows the cause to 1 thing.
#The 2 URLs, side by side
| Test URL | Production URL | |
|---|---|---|
| Path contains | /webhook-test/ |
/webhook/ |
| Registered when | You click Listen for Test Event | The workflow is Active |
| Lifetime | Until the editor stops waiting | Until you deactivate |
| Shows in Executions | No | Yes |
| Right for a real integration | No | Yes |
#1. Check which URL the sender is using
Look at the address configured in whatever is calling you, not the one in your notes. If it contains webhook-test, you've found the problem and the fix takes 30 seconds.
This is the cause maybe 3 times out of 5. The test URL is the one shown while you're building. It's right there when you're wiring things up, and nothing about it announces that it's temporary. Copying it is the obvious thing to do.
Swap it for the production URL, which the node shows on its other tab. Then move on.
#2. Make sure the workflow is actually Active
The production URL is registered when the workflow is activated and unregistered when it isn't. An inactive workflow's production URL returns 404 because, as far as the server is concerned, that route doesn't exist.
The toggle is in the top right of the editor. Saving is not activating. Saving is not activating, and that trips up a lot of people — a workflow can be saved, correct, and completely unreachable at the same time.
After switching it on, check the Executions tab. Production webhook calls appear there, which also gives you the fastest way to tell whether a request is arriving at all. If nothing appears, the request isn't reaching n8n and the problem is further upstream than the workflow.
#3. Match the HTTP method
The route is registered per method. A Webhook node expecting POST does not answer GET, and the error you get back says the webhook isn't registered — which reads like the path is wrong when the path is fine.
This catches 2 situations in particular. Opening the URL in a browser to "check it works" sends GET, and the 404 you get convinces you something is broken when nothing is. And a form or third-party service may send GET where you assumed POST.
Check the method on the node. Check what the sender actually sends. Make them agree.
#4. The URL n8n shows you may not be reachable
On a self-hosted instance behind a proxy, the address shown in the editor is built from configuration rather than discovered. If it shows localhost or an internal port, that's what gets copied into the sender, and the request never arrives.
Set the public address explicitly and restart n8n. The same reasoning as the container networking in the ECONNREFUSED guide applies: the address n8n knows about and the address the outside world can reach are 2 different things, and nothing reconciles them automatically.
#5. Check the path hasn't drifted
The path is part of the registration. Renaming it, or regenerating the id, changes the URL and leaves the old one unregistered. If somebody edited the node after the integration was configured, the sender is calling an address that genuinely no longer exists.
Compare the current URL on the node against the one in the sender, character for character. A trailing slash is enough to break it. A trailing slash difference is enough.
#6. A habit that prevents all of this
Copy the production URL first. Build second.
I now open the Webhook node, switch straight to the production tab, copy that address into whatever is calling it, and only then start wiring the workflow. The test URL still gets used while building, but it never reaches the sender, because the sender was configured before I had a chance to copy the wrong thing.
It sounds trivial. It removes the most common cause of this error completely, and it takes no extra time.
The second habit is activating before announcing. A workflow that is finished but switched off looks identical to a finished one from the editor, and the only place the difference shows is in a 404 somebody else sees. Check the toggle, fire one real request, confirm it lands in Executions, then tell people it is ready.
#Match your symptom to the fix
| What you're seeing | Likely cause | Go to |
|---|---|---|
| Works while listening, 404 after | Test URL in production use | 1 |
| Production URL 404s | Workflow not Active | 2 |
| 404 only from a browser | Method mismatch | 3 |
| URL shows localhost | Proxy configuration | 4 |
| Worked for weeks, then stopped | Path or node edited | 5 |
#When it's still 404
Call the production URL yourself, from outside your network, with the exact method the node expects:
curl -i -X POST https://your-n8n-host/webhook/my-hook -d '{}'
A 404 with the workflow Active points at the path. A connection error rather than a 404 means you're not reaching n8n at all, which is a proxy or DNS problem and has nothing to do with the workflow. A 200 from your test while the real sender still fails means one thing. The sender is configured differently from what you think. Ask it to log the exact URL it called, rather than trusting the field you looked at.
Frequently asked questions
Why does n8n have 2 URLs at all?
So you can iterate without running a live workflow. The test URL gives you the request data in the editor for building; the production one runs the saved workflow in the background. The cost is that they look nearly identical.
Does the test URL ever work when the workflow is active?
No. The test registration is for the editor session, and an active workflow serves the production route instead. Anything relying on the test URL will break.
Will deactivating lose queued calls?
Yes. An inactive workflow has no route, so calls during that window get a 404 and are gone unless the sender retries. Senders that don't retry simply drop the event.
Where do I see production webhook calls?
The Executions tab. If a call isn't listed there, it never reached the workflow, which immediately tells you to look at the network rather than the logic.
Written by
Vemuri Abhiram
Founder & Editor
Vemuri Abhiram is the founder and editor of Underread, where he writes about AI tools, software, and practical technology. He covers hands-on tool reviews, detailed guides, product comparisons, and emerging technologies.
You might also like
How to Fix “CUDA out of memory” When Running a Local LLM
Why a model that loaded yesterday runs out of VRAM today, and the 5 fixes in the order that costs you the least quality.
Vemuri Abhiram6 min read
How to Fix Cloudflare Error 524 on Long LLM Requests (Without an Enterprise Plan)
Why a slow model call dies at exactly 100 seconds behind Cloudflare, and why streaming fixes it when raising timeouts can't.
Vemuri Abhiram6 min read
How to Fix “operator does not exist: vector <=> vector” in Supabase and pgvector
Why a pgvector similarity search fails with error 42883 inside a Supabase function, and the fix with the fewest side effects.
Vemuri Abhiram5 min read
More from Underread, once a week.
The guides worth reading and the tests worth trusting. One email, every Friday.
No spam. Unsubscribe in one click.

