Lexicon of the most common Power Automate error codes: HTTP status, connector, and timeout errors with solutions.
A flow turns red, the run history shows an error code like 403 or an internal name like ActionFailed, and the error message itself doesn't really help you move forward. That's exactly where the searching begins: is it a problem with the connection, the input data, or the target service? Without a reference, every single failure costs valuable time, because you first have to find out what the code even stands for.
This lexicon summarizes the most important HTTP status codes and the typical Power Automate-specific error names, as documented in the official error code reference for cloud flows from Microsoft. For each error you get a short explanation of what it means, the most common causes behind it, and where to look first. Simply look up the code that's currently showing up in your run history.
Many connector errors in Power Automate are ultimately classic HTTP status codes returned by the API being called. Microsoft lists the most common ones in the documentation on connector integrity:
Good to know: errors in the 400 range are almost always caused by the request itself, while errors in the 500 range point to a problem on the called service's side and can usually only be resolved by waiting or retrying.
This group of errors occurs when the flow can no longer log in to a connected service.
According to Microsoft's documentation, it's worth switching production flows to a service principal connection, because it does not expire when an employee changes their password or leaves the company.
These errors come from the called service itself, not from Power Automate.
For all three errors, the same first step helps: open the failed run in the run history and compare the inputs with the format that's actually expected.
These errors show up either when saving or only at runtime, when an expression is evaluated.
coalesce() safeguard usually fixes this.This group occurs as soon as a flow exceeds time or volume limits.
Instead of googling every error individually, a fixed sequence helps, as also described by Microsoft in the guide Troubleshoot cloud flow errors:
If a flow at your company regularly gets stuck on the same error codes, or if you can't pin down the cause even with this reference, we'll take a look together as part of our Power Automate consulting. As digital employees, we set up connections, error handling, and retry logic properly once and document them so that you keep control over your flows afterward.
HTTP status codes like 401 or 404 come directly from the called service and describe why the specific request failed. Errors like ActionFailed, on the other hand, are a generic wrapper from Power Automate, behind which one of these HTTP codes is almost always hiding. Open the outputs of the failed action to find the actual status code behind it.
A 403 doesn't necessarily mean incorrect credentials, but rather missing permission for the specific operation. The most common causes are a data loss prevention policy blocking the connector in your environment, missing write access to the target resource, or an exceeded API rate limit. First check the permissions on the target resource, then the DLP policies in your environment.
Open the affected flow, select the failed run from the list of recent runs, and look for the red warning icon. You can expand the affected action, where the Inputs and Outputs sections show the exact status code and the original error message from the called service.
For one-off HTTP errors like a 429 or 502, simply retrying is usually enough, because the cause lies with the external service and is temporary. If the same error occurs repeatedly, however, it's worth setting up structured error handling in the flow, for example with retry policies and a defined fallback action, instead of manually restarting it every time.
The codes described here apply to cloud flows. For desktop flows that run on your PC, Microsoft maintains a separate reference titled Handling desktop flow errors, since different error sources such as UI elements or local applications play a role there.
Founder of NordFlux. Spent four years automating processes at enterprise scale at Dräger, and now brings that depth to the mid-market — pragmatic and with full data sovereignty.
Certifications
The 15 most important Power Automate expression functions such as concat, formatDateTime and coalesce, explained with German practical examples.
Does a guest user in Power Automate need their own license? The key answers based on official Microsoft documentation.
Why Power Automate approvals time out after 30 days and how the relay pattern of two flows reliably handles long-running approvals.
A lexicon helps you look things up, but it does not replace systematic error handling in daily operation. NordFlux takes over ongoing monitoring for your flows, so HTTP, connection, and timeout errors get caught before they stop a process.