Is GitHub down right now?

1. 1

   Read the live status above.

   githubstatus.com lists Git operations, API, Actions, Pages, Packages, Codespaces and Copilot separately: one can be down while the rest work.

2. 2

   Test Git directly.

   Run git ls-remote https://github.com/git/git.git. If it answers, GitHub is reachable over HTTPS from your machine.

3. 3

   Check your credentials.

   Expired personal access tokens, a missing SSH key on the agent, or SSO not authorized for your organization all fail like an outage.

4. 4

   Try HTTPS instead of SSH (or the reverse).

   Some networks block port 22. git over HTTPS uses port 443.

5. 5

   Check your network.

   Corporate proxies and firewalls often block raw.githubusercontent.com or the Copilot endpoints specifically.

Copilot can fail on its own while the rest of GitHub works. We check the Copilot API endpoint separately, and githubstatus.com lists Copilot and Copilot AI model providers as their own components: if either shows a problem, it appears in the official status box above.

If Copilot is up but not working for you: sign out and back in to GitHub in your editor, update the Copilot extension, check that your organization still allows Copilot for your account, and check whether a corporate proxy blocks the Copilot endpoints.

• Keep working locally. Git is distributed: commit locally and push when GitHub recovers.

• Pause deploy pipelines. Half-run Actions workflows can leave deployments in odd states. Rerun them once Actions is green.

• Do not rotate tokens in a panic. Authentication errors during an incident are usually GitHub-side; changing credentials adds a second problem.

• Cache what you depend on. Builds that download from raw.githubusercontent.com or GitHub releases on every run break during outages. Mirror or cache them.

Database and data-centre failover

In October 2018 a 43-second network cut between two facilities led to 24 hours of degraded service while GitHub protected data consistency.

Deployments and configuration

Many incidents follow a change that is rolled back once detected.

Dependencies on Azure and AI providers

Copilot depends on model providers, which is why GitHub lists them as a separate status component.

Load spikes

Heavy Actions usage and large events can slow queues and APIs.

The biggest GitHub incidents and what caused them. Durations are approximate and come from public reporting and official reviews.

• OCT 7, 2026

  GitHub outage confirmed after thousands of reports, recovered by early afternoon US Eastern time

  GitHub confirmed and investigated the incident on its status page. Source: Android Authority

• OCT 21, 2018

  24 hours and 11 minutes of degraded service: webhooks and Pages builds stalled

  Replacing failing optical equipment cut connectivity between the US East Coast hub and data center for 43 seconds, splitting database clusters. Source: GitHub post-incident analysis

Every minute our servers request GitHub's main endpoints, resolve github.com through Cloudflare, Google and gitHub's own nameservers, and read githubstatus.com. We combine that with one-tap user reports, filtered for duplicates and abuse: a spike needs at least 12 reports in 15 minutes from 8 or more different networks.

Built to avoid false alarms: a failed request is retried before it counts, a single bad check never says "down" on its own, and a check that fails for reasons on our side is shown as inconclusive rather than as an outage. Limits: we check from one location, so a regional problem may first show up in user reports. Reports store only the problem type, a country code and anonymous codes that change daily; never your IP address. Protego is independent and not affiliated with GitHub. Last reviewed October 26, 2026.