Need help fixing this n8n v2.x error (HTTP Request Tool + AI Agent):

Questions
1

I’m using HTTP Request Tool as a tool for an AI Agent in n8n v2.x, with credentials attached (HTTP Header Auth / API key).

I keep getting this error:

Problem in node ‘PartialExecutionToolExecutor’ Authorization failed – No credentials provided. Please login.

What I’ve already checked:

  • Tool node (not regular HTTP Request)

  • Authentication set to Generic Credential Type

  • HTTP Header Auth credential attached

  • API works fine via curl

  • Same workflow works in n8n v1.x, fails in v2.x

Question:
What is the correct way to fix this in n8n v2.x?
Is this expected when step-executing tools, or am I missing a required configuration?

Any guidance would be appreciated. Thanks!

2

Hey @Aye Welcome to the n8n community!

In n8n v2.x, the legacy HTTP Request Tool can behave differently when step‑executed as a tool than when run through the AI Agent, especially around URL/credential handling.

To work around this:

  • Use a regular HTTP Request node with your Header/Bearer credentials instead of the legacy tool, and let the AI Agent call that, or

  • Run the workflow from the AI Agent node (full execution) rather than step‑executing the tool; the partial PartialExecutionToolExecutor path isn’t documented to fully support credentials.

Sharing your exact workflow might help us narrow down this problem for you, let me know if this helps!

3

Screenshot (1553)
Screenshot (1553)1907×716 142 KB
THis is the exact workflow. I’m running into an “Invalid token” error in n8n, even though the authentication setup and token have been verified on the backend.

Error: Problem in node ‘PartialExecutionToolExecutor’
Error executing tool: Authorization failed – please check your credentials
Details: Invalid token

What’s confusing is that this setup was working perfectly before with no changes to the API or credentials.

Here’s what I’ve checked:

  • Backend authentication method and token validation are correctly configured

  • The token itself is valid and works when tested directly

  • Auth headers are correct and unchanged

  • Same setup previously executed without issues

This makes it seem like the issue isn’t the token itself, but something in the execution path that’s affecting how the token is being passed or resolved, especially when the tool runs through PartialExecutionToolExecutor.

Has anyone encountered a situation where:

  • A previously working setup suddenly reports “Invalid token”?

  • A valid token fails only in certain execution contexts?

  • Tool execution changes how credentials are injected?

Any insight into what could cause this would really help.

Thanks.

1 Like
4

This is expected behavior in n8n v2.x when an AI Tool is executed via PartialExecutionToolExecutor (step execution). In that execution path, credentials are not reliably loaded/injected, so the HTTP Request Tool ends up sending no auth header (or an unresolved one), which surfaces as “No credentials provided” or “Invalid token” even though the token is valid.

Fix / workaround i think will help you out:

  • Don’t use the legacy HTTP Request Tool for authenticated calls

  • Use a regular HTTP Request node inside a separate workflow

  • Call it via Execute Workflow Tool (or run the agent via full execution)

Alternative:

  • Manually set the Authorization header (env var / static token) instead of relying on credentials injection.

So the token isn’t actually bad, it’s the tool execution context in v2 that’s dropping credentials.

Let me know if this helps!