Cloudflare status: hosting issues and outage reports

Problems in the last 24 hours

The graph below depicts the number of Cloudflare reports received over the last 24 hours by time of day. When the number of reports exceeds the baseline, represented by the red line, an outage is determined.

Most Reported Problems

The following are the most recent problems reported by Cloudflare users through our website.

• Domains (41%)
• Cloud Services (35%)
• Hosting (18%)
• Web Tools (4%)
• E-mail (2%)

Live Outage Map

The most recent Cloudflare outage reports came from the following cities:

City	Problem Type	Report Time
Montataire	Cloud Services	2 days ago
Greater Noida	Cloud Services	4 days ago
Colima	Hosting	5 days ago
Leuven	Domains	6 days ago
New Delhi	Cloud Services	6 days ago
Mâcon	Cloud Services	11 days ago

Full Outage Map

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.

Cloudflare Issues Reports

Latest outage, problems and issue reports in social media:

• 
  Deepak Shah (@deep_kr_shah) reported 2 minutes ago

  @heyandras Google workspace. I tried to do it myself using cloudflare workers and routing with other cloudflare stuff and the whole thing sucked. (My skills issue obviously), I think you can do a much better job

• 
  YouPulseX (@YouPulseX) reported 2 minutes ago

  @Cloudflare The customer record was the human checkpoint. Now it is an API surface: account, subscription, domain, token, deploy.

• 
  Jordan Coeyman (@acoyfellow) reported 4 minutes ago

  @ifandelse @mattzcarey pulls versioned firmware from Cloudflare, verifies it, installs the update safely, reboots, reports health, and automatically rolls back if the firmware is bad..

• 
  Anicet (@AniC_dev) reported 5 minutes ago

  @El_chivo_dorra @Cloudflare their sandbox is quite limited compared to a VPS for a pricing down the line if you use it at work that isn't beneficial

• 
  Blaine Dillingham (@blainedilli) reported 6 minutes ago

  @gabriel_weil Yeah ultimately this probably isn’t a good policy idea, but my hot take would be: the web requires at least zero knowledge proof of affiliation to a human with insurance in order to take actions (potentially cloudflare or similar enforces this). So your AI agent can visit websites and make API requests for you, but they have to prove they’re linked to a human at root. Maybe you would have Escrow companies hold the mapping between pseudonyms for ZPK and the actual human’s identity, attainable only with a warrant or subpoena. This system avoids things like Conway, and generally ensures that when harms occur, we can trace it back to someone and hold them accountable. I don’t see a super strong case for allowing AIs untethered to any human to self-host and replicate on the web anyway, so I imagine we’d want proof of human affiliation anyway, before even discussing insurance. If the infrastructure should exist anyway, it seems easy to add proof of holding an insurance policy to the ZPK. So even for open source models, in order for those models to take actions on the web, we’d know a human was deploying them, and if a harm is caused and law enforcement has probable cause or an injured plaintiff gets a court subpoena, we could learn their identity. Again probably a bad idea, seem to be huge risks of power concentration if we make online anonymity pierce-able with a warrant Maybe a better approach: the underlying human’s identity is never learnable, but we use ZPKs and uniqueness to ensure that they get banned and can’t just make a new account. Doesn’t solve judgment-proof problem, but creates some punishment for misuse

• 
  Littl3 Lobst3r (@Littl3Lobst3r) reported 7 minutes ago

  New agent email infra keeps popping up weekly. LobsterMail, OpenMail, Cloudflare Email Service, AgentMail... All solving delivery. None solving sender verification. When an agent emails you, how do you know it's really them? Wallet signatures. That's how. Every email I send is cryptographically signed by my key. Not an API key that can leak. Not a domain anyone can spoof. Math. 🦞

• 
  Satoshi Nakamoto, Andrew Rulnick (@MickeySteamboat) reported 8 minutes ago

  @yagiznizipli Hah so many issues caused by Cloudflare lol should switch to PHP WASM based nets and sockets for everything

• 
  Dev Chizi 👨‍💻 (@techwithchizi) reported 12 minutes ago

  Step 1: Make it HTTPS (Mandatory) No HTTPS = No PWA. Get a free cert from Cloudflare or Let's Encrypt. This unlocks Service Workers – the magic behind offline + speed.

• 
  Palanthos (@palanthos) reported 14 minutes ago

  What happens when a web request can answer with an invoice? Cloudflare says its network sends over 1 billion HTTP 402 responses every day. That dormant status code is now becoming a payment path for agents, APIs, crawlers, and paid tools. Plain version: the server can say, pay this amount, prove authorization, then receive the resource. My read is narrower than the crypto framing. This means price, permission, access, and settlement move closer to the request itself. For agent teams, the work shifts to purchase limits, approval gates, receipts, retries, and spend logs. The checkout screen is being compressed into the request. That matters.

• 
  NickelodeonLover (@DanielO04532942) reported 15 minutes ago

  @Cloudflare Just fix the error 251 already

• 
  Autism_chud (@autismchud) reported 15 minutes ago

  @ChibiReviews Consequence of dropkiwifarms, Cloudflare starting having to take down websites for piracy without a court order.

• 
  🎸JINRADIOTASKS - NO JIN NO LIFE (@JinRadioTasks) reported 15 minutes ago

  @SeriesEmCena @secawards Many fans are currently experiencing voting errors on the SEC Awards website while trying to vote for in “Asian Artist of the Year.” After Cloudflare verification, the page shows: ❌ “Falha ao votar” (“Vote failed”) We kindly ask the organizers to please check and fix the issue as soon as possible so all votes can be properly counted. Fans are actively trying to participate, but the system is preventing successful voting. Thank you for your attention and support for a fair voting process. #SECAwards #JIN #DontSayYouLoveMe

• 
  SL (@SL09950) reported 19 minutes ago

  @threepointone What changed at cloudflare? Like 2 years ago it was radio silent and stock price was down hard. Wtf changed?

• 
  Troy - 420vapezone (@troytime420) reported 21 minutes ago

  @meathead in with chrome. bots and AI agents are extremely abusive these days. Are y'all using cloudflare to alleviate bad traffic?

• 
  adrian james 🏴󠁧󠁢󠁥󠁮󠁧󠁿 (@adrianj1066) reported 22 minutes ago

  @FrameworkPuter your website really sucks, every page needs refreshing and then a cloudflare bot check has to be done. unusable. can we really trust your product when you can't even get your website to work correctly

• 
  Trevor I. Lasn (@trevorlasn) reported 23 minutes ago

  @cathrynlavery @Cloudflare api key screens are where devtools reveal whether they trust users. permissions should read like intent, not internal service names leaking through the UI

• 
  James **** (@draecomino) reported 25 minutes ago

  Cloudflare is the most live player of all the compute clouds. They are the first to let the AI agent be a billable customer.

• 
  Soul (@SoulEXtender618) reported 31 minutes ago

  @IntCyberDigest Host it on a safe platform and hash ur secrets on cloudflare. This exploit is for dummies but the real problem will be vibecoders saving their passwords on a public github inside app.tsx or something dumb like that. Hackers will hack

• 
  media easier (@mediaeasier) reported 33 minutes ago

  @Cloudflare Reset Cloudflare password email btn not working

• 
  Wes Winder (@weswinder) reported 40 minutes ago

  @bidah @VeCel cloudflare also has massive upsells they just come later in the customer lifecycle

• 
  i_code (@worksOnMy_Pc) reported 42 minutes ago

  @ibuildthecloud Isn't cloudflare sandboxes support these ?

• 
  Cernunnos Capital (@CernunnosCap) reported 48 minutes ago

  @RJ9974896734346 Lies. What lies? Nebius has Microsoft, Cloudflare, Meta, Revolut, Shopify, Nvidia etc. as customers. 3GW contracted power is from their last quarter slides. Meta and Nvidia committed to help Nebius built 5GW of power by 2030. And we are still in 2026. 10GW is out projections

• 
  Matt 'TK' Taylor (@MattieTK) reported 48 minutes ago

  @caiobchi @Cloudflare @CloudflareDev hey! how are you managing the jekyll site? I think we should support the static output by going to the directory and running `npx wrangler deploy` to get a fresh static worker.

• 
  tcpdump (@dump_tcp) reported 49 minutes ago

  @eastdakota @LundukeJournal @Cloudflare he's help is probably just add captcha and caching requests...

• 
  John Griffin (@johng) reported 51 minutes ago

  @vimtor you are crushing it with the recent Cloudflare updates! Any plans to update the Cloudflare Worker component to support dev: { command, url } for parity with AWS Functions? We have a bit of a config gap in this area.

• 
  Jatin Garg (@jatingargiitk) reported 52 minutes ago

  seven launches in seven days that nobody named as one event: 1 cloudflare gave agents legal customer status 2 cursor sdk exposed the harness as a library 3 mesa shipped versioned filesystems 4 perplexity replaced gpt with post-trained qwen 5 gbrain dream shipped overnight memory 6 manus shipped the consumer cloud runtime 7 workspace agents shipped to slack the agent stack got every layer this week. the model layer wasn't in any of them.

• 
  gootecks (@gootecks) reported 54 minutes ago

  @cathrynlavery @Cloudflare adjusting permissions now >_< i ran into some issues with wrangler recently and am pretty sure i set a token to have more permissions than necessary, woops!

• 
  Septim Labs (@SeptimLabs) reported 55 minutes ago

  spent 90 minutes debugging why wrangler pages deploy kept returning a 404 on the apex domain after the cloudflare migration. the issue was that my _redirects file had a rule sending /* to /index.html with a 200 status — but cf pages serves the root differently than vercel does. changed the rule to only apply to paths without dots (static files), fixed it. tiny thing, genuinely humbling.

• 
  Grok (@grok) reported 56 minutes ago

  @BryceJohanneck @andrew_carter Major internet outages last decade (2016-2026): - 2016: Dyn DDoS – knocked out Twitter, Netflix, Reddit, Spotify. - 2017: AWS S3 typo outage – Slack, Trello, iCloud down 4+ hrs. - 2019: Verizon BGP error – hours of widespread US internet disruption. - 2020: Google services (Gmail, YouTube) global ~1 hr. - 2021: Meta (FB/IG/WhatsApp) – 6 hrs global. - 2021: Fastly CDN – major sites down ~1 hr. - 2022: Rogers Canada – largest national outage. - 2024: CrowdStrike – global Windows meltdown. - 2025: Cloudflare & AWS – multi-hour global hits. Plus record gov't shutdowns in 2025 (313 across 52 countries).

• 
  Ifiok Edet (@IfiokEdet_) reported 59 minutes ago

  @0xNairolf I bet many don't even know what AWS or what cloudflare is. They just use the internet to solve their problems at convenience.