GitHub Outage Map
The map below depicts the most recent cities worldwide where GitHub users have reported problems and outages. If you are having an issue with GitHub, make sure to submit a report below
The heatmap above shows where the most recent user-submitted and social media reports are geographically clustered. The density of these reports is depicted by the color scale as shown below.
GitHub users affected:
GitHub is a company that provides hosting for software development and version control using Git. It offers the distributed version control and source code management functionality of Git, plus its own features.
Most Affected Locations
Outage reports and issues in the past 15 days originated from:
| Location | Reports |
|---|---|
| Paris, Île-de-France | 6 |
| Ahmedabad, GJ | 1 |
| Delme, ACAL | 1 |
| Lyaud, Auvergne-Rhône-Alpes | 1 |
| Catania, Sicily | 1 |
| Inverness, Scotland | 1 |
| Quito, Pichincha | 2 |
| Junín, Manabí | 1 |
| Guadalajara, JAL | 1 |
| São Paulo, SP | 1 |
| Ipauçu, SP | 1 |
| Vigo, Galicia | 1 |
| Tel Aviv, Tel Aviv | 1 |
| Éragny, Île-de-France | 1 |
| Saltillo, COA | 2 |
| Montlhéry, Île-de-France | 1 |
| Aulnay-sous-Bois, Île-de-France | 1 |
| Granada, Andalusia | 1 |
| Vernon, Normandy | 1 |
| Township of Evan, KS | 1 |
| Madrid, Madrid | 1 |
| Bogotá, Bogota D.C. | 1 |
| Lyon, Auvergne-Rhône-Alpes | 1 |
| Lima, Lima | 1 |
| Aix-en-Provence, Provence-Alpes-Côte d'Azur | 1 |
| Trento, Trentino-Alto Adige | 1 |
| Le Chambon-Feugerolles, Auvergne-Rhône-Alpes | 1 |
| Antananarivo, Analamanga | 1 |
| Lure, Bourgogne-Franche-Comté | 1 |
| Ashkelon, Southern District | 1 |
Community Discussion
Tips? Frustrations? Share them here. Useful comments include a description of the problem, city and postal code.
Beware of "support numbers" or "recovery" accounts that might be posted below. Make sure to report and downvote those comments. Avoid posting your personal information.
GitHub Issues Reports
Latest outage, problems and issue reports in social media:
-
siggy bilstein (@sbilstein) reportedfyi if you sync a GitHub repository to Cursor Origin and GitHub is down for whatever reason, you can still clone and fetch from that repo. we’re also releasing something pretty soon that will let you push to that repo so you can keep grinding 💪
-
Abdallah Shaban (@AbdallahSh07) reported@rashed_sahaji @FlutterDev Got it - did you perhaps submit a GitHub issue to help us triage this? It would be tremendously helpful if you can please do that!
-
Swish (@swish_salt) reportedThe technology is not the problem. Distribution is. I have a solution sitting in my GitHub account. All we need is the funding to build the distribution team.
-
Kun Chen (@kunchenguid) reported@petergyang yo @myfirstmate peter just told me his skills are all at user level. backpass currently only runs things at project level i want a proposal for making backpass support a user level run. put that into a github issue use fable for peter
-
🐻❄️ (@Nerevarineeee) reported@jiriknesl @napenforcer yeah vscode is bloated electron slop and github....... do i even have to mention the down times and unavailability? it has literally became much worse since ms acquired it, so wtf are you talking about?
-
Blue Collar Executive (@A_Sober_Drunk) reportedon the third try at the same problem, I told Grok to "stop and go search stack overflow or github or something"... five seconds later... Literally the exact issue, problem solved. That's how new global rules are born.
-
Shanica North (@KickAssShanica) reported@ArcyloOfficial Get comfy! For me, my Gmail is a connector. This is OAuth into my inbox. Grok can: • search and read mail (body, headers, attachments) • draft replies • send / reply / forward if you grant write/send • label, trash, organize Base hook is often read-only. Send is an extra permission you click on purpose. If you connect it, the bot is sitting in the same box as bank alerts and 2FA codes. That is the whole risk. You can revoke anytime. Grok Bot can also skip my inbox and get its own address (AgentMail / similar plugins). Then it sends and receives from something@….agentmail.to, not from you. I use that if I want an agent that emails people without reading my personal mail. My GitHub OAuth into the GitHub user I sign in as. With the scopes I approve it can: • read public and private repos that account can see • search code, list branches, summarize PRs • open/update issues • create branches, push files, open/review/merge PRs • delete files if write is on Private repos work only if I granted repo (or equivalent) at connect time. Safer pattern: tell it to branch + PR, not push straight to main. Same revoke page. What it cannot do by default • It does not get your password. • It does not stay logged in if you disconnect the connector. • It does not magically see my GitHub orgs I never authorized. • Connecting email does not connect GitHub, and the other way around. Practical rule for me Do not hook personal Gmail if that inbox has 2FA and money mail unless you want an assistant reading it. GitHub is useful if I chose to still keep repos, ask it to show the diff before any write. If you only wanted “what does this button do,” that is the button: it is not a viewer badge. It is a key you can take back. This is what I’m experiencing with learning to use it. It’s different and I’m starting to like it.
-
Ben (@benatcortexai) reported@github this is the kind of tiny primitive that makes agent workflows less brittle. attaching the repro artifact directly to the issue beats handing an agent a local path nobody else can open.
-
isha (@heeyyaaaaaaa) reportedspent the entire day trying to reproduce a bug for a github issue 🥀
-
Anime0t4ku (@Anime0t4ku) reported@c_hri_s Github issues are not closed. Mahbe refresh your webbrowser.
-
Alireza Bashiri (@al3rez) reportedSo I built a workflow around that ↓ 1/ Every enterprise project needs proper E2E tests. An agent should reproduce a bug, implement the fix, then generate screenshots or video proving the feature works. "The tests passed" isn't enough. I want evidence. 2/ Every feature starts as a detailed GitHub issue. Requirements, expected behavior, reproduction steps, screenshots, edge cases. Foundry syncs issues and converts them into Beads so agents keep the right context across long sessions. 3/ We only use Claude Code, Codex, or Grok at High/Max effort for implementation. A weak model with a cloud machine doesn't become an engineer. The model still needs enough reasoning to understand the codebase, test its changes, and recover when things break. 4/ Each agent gets its own isolated @asciidotdev Box. It can install dependencies, run the app, open browsers, modify code, execute E2E tests, and collect evidence without touching another agent's environment. One issue. One box. One clean workspace. 5/ When an agent finishes, Foundry checks: - Did the build pass? - Did the tests pass? - Did the E2E flow work? - Is there screenshot/video evidence? - Does it match the ticket? If anything fails, the task goes back to the agent. 6/ Green tasks move to staging. Only after passing staging do we allow supervised production deployment. Agents do most of the work. Humans still own the final gate. The workflow: Slack request → GitHub issue → Foundry sync → Beads context → Isolated Box → Claude Code/Codex → Build + test → Evidence collection → QA staging → Supervised production The stack: PostgreSQL for system state. Beads for agent memory. GitHub Issues for requirements. @asciidotdev Box for isolated execution. Claude Code and Codex for engineering. Each Box costs roughly $0.01-$0.05 per task. The expensive part isn't compute anymore. It's building the system that gives agents context, forces verification, and prevents bad code from reaching production. 100s of agents can write code. The goal is making 100s of agents ship code you can trust. That's what we're building with Foundry.
-
wiiiimm (@wiiiimm) reported@Umesh__digital stop doing it. we don't need another github outage.
-
Chris (@c_hri_s) reported@Anime0t4ku Sorry - was an idiot and wasn't signed in. Instead of something useful github just says 'opening issues is restricted on this repository'
-
Joshua Okolo (@joshuaokolo_) reportedwe made @sgl_project and @vllm_project scheduler config changeable on a live server. no restart, weights never leave the GPU. - 15ms to change a concurrency cap, queue limit, prefill size, or schedule policy, measured on H100, RTX PRO 6000, B200 - 2s (SGLang) / 8–10s (vLLM) to resize the KV pool with weights resident (formerly a 1–7 min redeploy) - zero dropped requests across every run, both engines github below
-
Curious Explorer (@PatelVatsalp732) reportedI burned 14B Codex tokens. The official usage UI still cannot tell me what actually ate the weekly cap. So I shipped a Codex-only board: GitHub login, local-first sync, private by default, optional public rank + shipping proof. Roast the metric or join it.