A8gent
All articles

Technical

Unable to sign without access token: 4 Causes and the Fix for Each in n8n

Your OAuth2 credential says it is connected, but when a Google node runs or loads a list, n8n has no access token to send. The big wave of reports came from an n8n 1.74 bug fixed in 1.75.0. The rest come from proxies, an unfinished sign-in or a callback that never landed.

A8gentAutomation practiceNovember 17, 20269 min read

Short answer: "Unable to sign without access token" means the OAuth2 credential has stored token data but no access token in it, so n8n cannot sign the request. Most cases came from an n8n 1.74.x bug, fixed in 1.75.0. Upgrade, set N8N_PROXY_HOPS=1 if n8n is behind a reverse proxy, restart, then open the credential and sign in again. If it still fails, delete the credential and create it fresh.

Contents
  1. The Short Answer
  2. What the Message Means
  3. Cause 1: The n8n 1.74 OAuth Credential Bug
  4. Cause 2: n8n Is Behind a Proxy Without N8N_PROXY_HOPS
  5. Cause 3: The Credential Still Holds Bad Token Data
  6. Cause 4: The Sign-In Never Finished
  7. Still Broken? Check These
  8. How to Stop It Coming Back
  9. Sources
  10. FAQ

The Short Answer

The error appears when a node that uses an OAuth2 credential runs, or when the node tries to load a list such as your spreadsheets or calendars. It can be printed with an error class in front. Both of these appear in the reports:

NodeApiError: Unable to sign without access token
NodeOperationError: Unable to sign without access token

The credential screen often says the account is connected. That is what makes it confusing. n8n has a record for the connection, but the access token inside it is missing.

The fix that worked for most people in the main forum thread, which has passed 13,000 views:

  1. Upgrade n8n to 1.75.0 or later.
  2. If n8n runs behind a reverse proxy or Cloudflare, set N8N_PROXY_HOPS=1 and restart.
  3. Open the credential and sign in with Google again.
  4. If the node still fails, delete the credential, create a new one and select it in the node.

What the Message Means

The string comes from n8n's own OAuth2 client library. In client-oauth2-token.ts, the sign method adds the access token to an outgoing request, and its first line is:

if (!this.accessToken) {
  throw new Error('Unable to sign without access token');
}

So the request never reaches Google. The failure is inside n8n, before any call is made, which is why Google's console shows nothing for it.

There is a related message worth separating. In the current request helper, an authorization code credential with no token data at all fails earlier with OAuth credentials not connected. "Unable to sign without access token" means token data exists but holds no usable access token. That points at a credential that was damaged, or a sign-in that only partly completed.

SymptomCauseFixHow to confirm
Started after an update to 1.74.x, many credentials at onceBug that blanked OAuth token dataUpgrade to 1.75.0 or later, reconnectVersion in Settings or the Docker image tag
Self-hosted behind Cloudflare, nginx or another proxyN8N_PROXY_HOPS not setSet it to 1, restart, reconnectVariable missing from the container environment
Reconnecting says success, node still failsCredential still holds bad dataDelete and recreate the credentialA brand new credential works
Sign-in never really finishedConsent blocked or callback failedFix the Google side, sign in againGoogle showed an error page during sign-in

Cause 1: The n8n 1.74 OAuth Credential Bug

The thread exploded in mid January 2025. Users on 1.74.1, 1.74.2 and 1.74.3 reported that OAuth2 connections to Google stopped working after updating, while older installs kept working. One user saw it on nearly every OAuth2 credential at once.

n8n staff linked the cause in the thread: PR #12563, "Don't break oauth credentials when updating them and allow fixing broken oauth credentials by repeating the authorization flow". Its description names two bugs. Updating an OAuth credential saved a blank value for the token data, and signing in again did not repair it. The release notes list the fix in n8n 1.75.0, published 15 January 2025.

How to confirm it: check your version. On Docker, look at the image tag. Using n8nio/n8n without a version tag pulls whatever is current, which is how several users landed on 1.74.x without meaning to.

The fix: upgrade, then sign in again so the repaired code writes fresh token data. With Docker Compose, pin a version you have checked rather than relying on the default tag:

services:
  n8n:
    image: docker.n8n.io/n8nio/n8n:1.75.2

Use the current release for a new install. The point is to know which version you run. Rolling back to 1.73.1 did not fix every user, because a credential already blanked stays blank until it is reauthorized or recreated.

Cause 2: n8n Is Behind a Proxy Without N8N_PROXY_HOPS

The first fix posted in the thread came from a user behind Cloudflare. They turned off Cloudflare proxying, set N8N_PROXY_HOPS to 1, turned the proxy back on, and OAuth2 worked. Others on Docker and Kubernetes confirmed the same, usually together with an upgrade and a reconnect.

n8n's documentation does not tie this setting to this error, and we could not find a documented mechanism, so treat it as a confirmed workaround rather than an explained one. What the reverse proxy documentation does say is that behind a proxy you should set N8N_PROXY_HOPS to 1 and have the last proxy pass X-Forwarded-For, X-Forwarded-Host and X-Forwarded-Proto. So it is correct configuration either way.

How to confirm it:

docker compose exec n8n env | grep N8N_PROXY_HOPS

The fix:

services:
  n8n:
    environment:
      - N8N_PROXY_HOPS=1

Recreate the container with docker compose up -d --force-recreate n8n, then sign in to the credential again. One user only succeeded after doing all three: upgrade, set the variable, and delete and recreate the credential.

Cause 3: The Credential Still Holds Bad Token Data

Reconnecting reports success, yet the node fails with the same message. Several posters in the thread fixed it only by deleting the credential and creating a new one. One added a detail worth repeating: creating a second credential while keeping the original did not work for them. Deleting the original first did.

How to confirm it: create a fresh credential with the same Client ID and Client Secret and test one node with it. If the new one works, the old one is damaged.

The fix:

  1. Note which workflows and nodes use the credential.
  2. Delete it.
  3. Create a new credential of the same type, paste the Client ID and Client Secret, and select Sign in with Google.
  4. Open each node that used the old one and select the new credential. Nodes do not switch automatically.

Your Google Cloud client does not need to change. The OAuth Redirect URL is the same for every credential on an instance.

Cause 4: The Sign-In Never Finished

Two later replies in the thread describe this. One user had saved the Client ID and Secret but had not selected Sign in with Google at the bottom of the credential screen. Another found their app's audience was not set up, so Google never issued the tokens. Adding their own address as a test user and signing in again fixed it.

The n8n Google OAuth2 guide covers both. Its last step is to select Sign in with Google. And an External app in Testing mode only lets accounts listed as test users finish the flow, which you add on the Audience page in Google Cloud.

How to confirm it: sign in again and watch every screen. An "Access blocked" page, a redirect_uri_mismatch, or a browser that ends up on a localhost address all mean the token never reached n8n.

The fix depends on which screen failed:

Still Broken? Check These

  1. n8n is on 1.75.0 or later.
  2. N8N_PROXY_HOPS=1 is set if anything sits between the internet and n8n.
  3. The OAuth Redirect URL in the credential shows your public https address, not localhost or an IP.
  4. The Google APIs the node calls are enabled in the same Google Cloud project. The n8n docs note that Google Docs, Sheets and Slides also need the Google Drive API. One user in the thread fixed their Sheets error by enabling Drive.
  5. Your Google account is a test user, or the app is published.
  6. The credential was deleted and recreated, and every node points at the new one.

Two neighbouring messages are different problems. If a credential worked for a week and then failed, the n8n docs explain that for an External app in Testing, consent and tokens expire after seven days. The current n8n source reports a revoked or expired refresh as The credential "name" needs to be reconnected. Reconnect it, or publish the app to stop the seven day cycle.

How to Stop It Coming Back

Pin the n8n version. Upgrade on purpose, after reading the release notes, not because a container restarted on a floating tag. This error hit people who did not know they had upgraded.

Test one OAuth node after each upgrade. Run a single Google node by hand before trusting scheduled workflows.

Set the proxy variables once, in the deployment file. N8N_PROXY_HOPS, the editor URL and the webhook URL belong in version control with the rest of the setup, so a rebuild cannot lose them.

Alert on failures. In Workflow Settings, set an error workflow that starts with the Error Trigger node, so a credential failure reaches you before a client notices. See error handling.

A credential that breaks quietly is one of the most common ways a client automation stops. The n8n AI Agents course covers building agents with tools and memory, triggers and webhooks, error handling and human fallback, and deploying on your own server or n8n Cloud. For a working Google example, see our n8n AI agent with Google Sheets, and for the hosting trade-offs, the true cost of self-hosting n8n.

Sources

Checked against n8n's source, release notes, documentation and community forum on 17 November 2026.

FAQ

What does "Unable to sign without access token" mean in n8n?
n8n tried to add the OAuth2 access token to a request and the credential had none. The message is thrown inside n8n's OAuth2 client before anything is sent to Google, so it points at the stored credential, not at the API.
Why does the credential say connected when the node fails?
The credential has a token record, but the access token in it is missing or blank. On n8n 1.74.x, updating an OAuth credential could save blank token data. Upgrade to 1.75.0 or later and sign in again, or recreate the credential.
Which n8n version fixed it?
n8n 1.75.0, released on 15 January 2025, includes PR #12563, which stops updates from breaking OAuth credentials and lets you repair a broken one by repeating the authorization flow.
Does N8N_PROXY_HOPS fix this error?
Several self-hosted users behind Cloudflare or another proxy fixed it by setting N8N_PROXY_HOPS=1 and reconnecting. n8n does not document a direct link, but its reverse proxy guide asks for that setting behind a proxy, so set it either way.
Should I reconnect or recreate the credential?
Try signing in again first. If the node still fails, delete the credential, create a new one with the same Client ID and Secret, and select it in every node that used the old one. One user found a second credential did not work until the original was deleted.
Does this happen on n8n Cloud?
Most reports in the thread are self-hosted, but one n8n Cloud user on 1.74.3 asked for help there too. The 1.75.0 fix is in n8n itself, so it covers both. The proxy setting only matters for self-hosted instances.
My Google credential stopped after seven days. Is that the same error?
No. For an External Google app in Testing mode, consent and tokens expire after seven days. Current n8n reports that as a credential that needs to be reconnected. Reconnect it, or publish the app.

If this resonated

Module 02 is the long version of this post.

Eval set, refusal pattern, kill switch. Build it in a weekend.

All postsLast verified 2026-11-17

Done for you

Want this built for your business?

Tell us the workflow. We scope, build, and ship the agent with guardrails and the numbers to prove it worked. The scoping call is free.

Free scoping call. You own the code.

Request a scoping call