A little context
behind the answer.

Two signals: what GitHub reports, and whether an unusual number of people are asking the same question.

The three answers

Yes: GitHub has an unresolved incident explicitly affecting Actions. This includes incidents being investigated, identified, or monitored, even if some workflows still run.
Probably: GitHub reports a disruption without an active Actions incident, or visitor interest has risen sharply. When interest is the only signal, GitHub has not confirmed an incident.
No: GitHub reports Actions operational and we see no qualifying surge. This is not a guarantee that every workflow is working.

We use the Actions component in GitHub’s public status feed. Incidents for other products and future scheduled maintenance do not change the answer.

How visitor interest works

We look for at least 50 eligible visitors in 5 minutes, and at least 5 times the preceding hour’s usual five-minute count. Both must hold for 2 consecutive minute checks.

The baseline needs a complete hour of observations plus the current five-minute window. Once interest is elevated, we hold the baseline steady and wait for ten minutes below the threshold before clearing the signal.

Repeated visits are deduplicated for 30 minutes. Known bots, automatic refreshes, and API or MCP polling are excluded. These are approximate visitor counts: shared networks may undercount, and a popular link can still cause a false alarm. Interest alone never produces “Yes.”

When a source is unavailable

We normally check GitHub every 30 seconds while someone is using the site. If the feed fails, we retain the last answer and its original timestamp. An old result is clearly marked, including whenever its check time is more than five minutes ago.

If we have never received a successful result, the answer is “Unable to verify.” If visitor data is unavailable, we continue to show what GitHub reports and label visitor interest unavailable.

A small privacy footprint

We do not use advertising trackers or analytics cookies. To limit duplicate submissions, the server temporarily uses a one-way keyed hash of the request’s network address and browser information. Raw IP addresses and browser strings are not saved in our application’s visitor store or logs.

Visitor identifiers expire after 30 minutes; aggregate minute counts are kept briefly for comparison. Hosting providers process request metadata to serve and protect the site. Agent requests contain no repository or account information.

An independent check