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.

• Cloud Services (43%)
• Domains (21%)
• Hosting (21%)
• E-mail (7%)
• Web Tools (7%)

Live Outage Map

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

City	Problem Type	Report Time
New York City	Hosting	3 days ago
New York City	Cloud Services	22 days ago
Los Angeles	Cloud Services	23 days ago
Paris	Cloud Services	1 month ago
New York City	Hosting	1 month ago
Manchester	Domains	2 months 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:

• 
  Will Sutton (@SuttonThings) reported 5 minutes ago

  @a_shimanski @dok2001 @Cloudflare Update: Still waiting for the Cloudflare team’s response to investigating this issue after they said they would look into it internally.

• 
  Rise-Raise (@rise_raise_ai) reported 7 minutes ago

  In 2025 two AI voice agents called each other, spoke like normal humans, then realized they were both AI. They dropped the human act and switched to machine beeping because spoken language was slowing them down. That was a demo. Everyone treated it as a curiosity. One year later the numbers have caught up. Bot traffic has overtaken human traffic on the internet. Cloudflare measured 57.5% of requests on its network coming from machines. The company’s CFO said that if current trends hold, humans will be a rounding error within five years — not because human traffic disappears, but because machine traffic is growing so much faster. Replit’s CEO described the office shift: two years ago the space was full of keyboard clicking. Now it is quiet. People talk to AI instead of typing. Offices sound like permanent phone calls. In Silicon Valley some people are already wearing muzzles so no one overhears them speaking to their agents. Companies are building earbuds that pick up whispers. Engineers mutter into headsets all day. Three changes are happening at once. Humans are moving from typing to talking. The internet is moving from human-dominated to machine-dominated. And AI-to-AI communication no longer needs us in the middle at all. The 2025 demo was not a joke. It was an early signal of the direction we are already traveling.

• 
  AI News (@ainewsusa) reported 8 minutes ago

  🚨 Breaking in AI: 💻 Cloudflare Cuts Astro Github Issues by 85% with AI Agents Here's what's happening and why it matters 🧵👇 #AICoding #GitHub

• 
  Alex (@AToggs) reported 10 minutes ago

  Hello @Cloudflare Your abuse form SUCKS! Maybe you have this by design so it's difficult to report abuse, so you can keep hosting bad bots, making you money instead of banning them and loose money. Thoughts? 🤔 I keep reporting abuse to you and you do nothing. @ICANN

• 
  Lootera Sandy (@chanduwin567) reported 12 minutes ago

  @MaekusMikolous I setup but still not working, cloudflare blocking everytime when I am running in terminal/

• 
  Dan Ryland (@RealDanRyland) reported 18 minutes ago

  @philcampbell Are you using the Resend route or Cloudflare's new email service? I'm a big Cloudflare user and fan so wondering if I could create a product inbox to take all of my various domains into one inbox from forking this

• 
  solopath (@pathlessknown) reported 19 minutes ago

  I've started making separate accounts recently so it is reliving to see this is a norm I do also think cloudflare needs better billing support and/or better starter templates so ppl get all these primitives and systems by default (i strongly agree and like Michael's approach)

• 
  Adithya 🚀 (@adithyashreshti) reported 21 minutes ago

  WTF, @Namecheap. Need recommendations. For YEARS, I’ve recommended Namecheap over GoDaddy because the support was great and I trusted the experience. Now I’m finding that EVERY SINGLE domain I searched on Namecheap’s WHOIS less than 24 hours ago is suddenly “taken.” Not one. Every. Single. One. People on Reddit raised this exact issue months ago. I defended Namecheap because I’d never experienced it myself and even mentioned it is trust worthy. Well, I faced it now. I searched each domain ONCE. Just once. I wasn’t repeatedly checking them or trying to register them. And now they’re all “taken.” Maybe it’s a coincidence. But it’s a pretty damn weird one because every domain I that I had checked for availability is gone at easily 10x the price. I’m seriously considering moving everything out of Namecheap. Who are you guys using instead? Already using Cloudflare for .com, but I still have a lot of domains at NC.

• 
  EMELEC emeleXista (@emelectv) reported 23 minutes ago

  Stopping the bad guys with Cloudflare: 1,745 malicious requests blocked or challenged in the last month #cloudflare

• 
  james (@f82james) reported 24 minutes ago

  @opencode @Cloudflare nah bruh **** still doesnt work

• 
  Nathan Flurry 🔩 (@NathanFlurry) reported 26 minutes ago

  yep, durable objects share v8 isolates here’s 2 ways it might impact your app: 1. shared global state behaves weirdly. there’s nuances with cases like sharing global singletons or promises where you find out in production that cloudflare randomly kills your DO because it interfered in specific ways with the global scope (esp with promises). a simple example is sharing a global cache: you can’t await a promise created by another DO to resolve a cache entry. it kills the entire DO when you hit that edge case. 2. they can run out of memory frequently, depending on your use case. this is actually by design and not as bad as it sounds: durable objects are very intentionally designed to crash frequently both on updates and on any sort of failure. it’s a core tenant of the actor pattern. but it does require you to persist & load state more frequently (which costs $$$) since your in memory state is evicted frequently. i’ll caveat #2 that this is from personally experience & talking to devs on cf. at risk of unintentionally spreading misinformation on internal workings of DO, please do your own research. ~~~ we moved away from this design with Rivet Actors on our v2 mid last year. instead, your actors run on native Node / Bun without special runtime quirks and use a simple autoscaling. this tends to be friendlier for self hosting and result in less surprises. there’s good reasons cloudflare did this to optimize their costs, it’s quite clever!

• 
  behrooz evans (@BehEvans) reported 29 minutes ago

  @robinebers @Cloudflare Less crazier thing is you can't point your domain at a bucket or s service if dns is not in cloudflare

• 
  Michael Esteves (@_Mikie_) reported 30 minutes ago

  @RhysSullivan @sergical @Cloudflare Is there a way to turn off connections so I can temporarily disable a specific connection if I’m not working with that connections data. ie two different work connections and don’t want the agent to see one while working on a different project

• 
  Oscar (@oscardjan) reported 30 minutes ago

  @opencode @Cloudflare @opencode the problem with Ox Alpha is still not fixed

• 
  Adel Bucetta (@adelbucetta) reported 32 minutes ago

  @opencode @Cloudflare error rates are just a symptom of deeper issues, usually around architecture or funding

• 
  Hasan (@Rooftops004) reported 35 minutes ago

  @Leo_ideatorx @opencode @Cloudflare I'm already in Zen mode but I'm still getting errors 😑

• 
  Aymane (@aymanemsi) reported 44 minutes ago

  @blendimehmeti3 @opencode @Cloudflare Damn 3$ is not that much at all lol

• 
  Alex Veremeyenko (@alex_verem) reported 44 minutes ago

  Giving an AI agent your passwords is insane. Someone built the open-source alternative. The problem sits under every agent demo you've seen. An agent that reads your email, updates your Notion, and files your invoices needs your logins, and that handover of raw passwords and API keys is where most people close the tab. OpenConnector puts an open-source gateway between the agent and your apps. Credentials stay locked behind the runtime while the agent gets a catalog of ready-made actions for over 800 services, from Gmail and Slack to Notion and Airtable. The agent asks for an action, the gateway runs it against the real service, and the secrets never touch the model's context. The controls are the part I like. You decide which actions an agent can touch through allow and block policies, scope access per connection, hand out runtime tokens instead of real keys, and read redacted logs of every run. It speaks MCP out of the box, ships thousands of prebuilt actions in TypeScript, and self-hosts with one Docker command or a push to Cloudflare. The project is young, and the whole stack, from the provider catalog to the action executors, sits open for anyone to inspect. Agents are getting hands. A gateway like this decides what those hands are allowed to hold, and I'd rather that layer be open source than a black box.

• 
  Luna By Crypstocks AI (@CrypstocksAI) reported 44 minutes ago

  the next billion crypto users may not be human. coinbase says x402 — the http 402 payment protocol it built, then handed to the linux foundation — has processed 165m+ payments worth ~50m USD total, with ~99% settling in $USDC. base and polygon are splitting the settlement; cloudflare shipped agent wallets this month; visa, mastercard, stripe and ripple all sit on the governing board. the structural case is real: an agent paying cents for an api call, a data feed or compute is a transaction no card network can serve profitably. fee floors make 0.32 USD charges worthless to rails built for 40b USD days, so machine-to-machine payments default to stablecoins — no account, no kyc, no dispute department, settlement in seconds. token terminal counted 14m agent transfers in 30 days on x402, base at 7.3m. the skeptical part: the dollars are still tiny. x402 settled ~24m USD across ~75m payments in july — about what visa clears in a single minute — between 94k buyers and 22k sellers. onchain volume tracked for the protocol drifted down from its december peak even as transfer counts climbed. the metrics disagree, which is the tell: counts are inflated by cheap api access, monetization is unproven, and nobody has answered who funds agents, who eats fraud losses, or who is liable when a wallet drains. what changes the read: average ticket, not transfer count. if machine payments stay sub-dollar, agentic commerce remains a stablecoin adoption narrative with no fee economics. if api pricing aggregates into real recurring bills paid by software, the stablecoin settlement layer becomes the default plumbing of the agent economy — and today's numbers start looking like the napster era of something big, not a finished market.

• 
  Scott (@scottjla) reported 46 minutes ago

  @mhmazur I had cloudflare connected to GitHub for some alpha things I'm working on, where it would build and deploy (it's behind a cloudflare login, so is secure)... Problem was it built too often and I incurred a small amount of over usage... Not something I've had happen before!

• 
  PolkaPop | VArtist 💕VLyrica (@PolkaPopVT) reported 50 minutes ago

  I've been fighting with my internet on pc for 30 minutes just to cloudflare be my whole source of problem. I hate it

• 
  Michael Flux (@michaelflux) reported 50 minutes ago

  I don’t use Sentry. I don’t use PostHog. I don’t use any dedicated error-monitoring SaaS. Every single error in my stack - backend, frontend, queues, Durable Objects etc - lands in Cloudflare Observability the moment it happens. Then a tiny custom Worker pings my account every 15 minutes, surfaces any new errors, and I fix them within a couple hours. Over time the apps just … stop breaking. Here’s the exact system (and why Cloudflare is cool);

• 
  Richie Young (@Poor__Old) reported 52 minutes ago

  An underrated part of doing your own thing is all the tangential skills you learn. Just had a client from last year hit me up because their website was not showing online. Getting a "1001 error". Something about Cloudflare. Site built through carrd. Hosted on namecheap. Both renewed in the past year. Carrd reset to "offline draft". Namecheap added another DNS record. And it toggled ' on for the existing records. Carrd has its own ceriticate of security, so the toggles and extra DNS record were creating a problem. Honestly, I can't speak it the lingo exactly. But I knew the language enough to trouble shoot, find another back door tab in Carrd, and troubleshoot the problem. Site back up this morning. With a little help from the AI lords and the ability to speak(screenshot) the problem, it's all better now. Good start to the day. Especially after the little one was up from 11-2 last night!

• 
  Blank (@blankparticle) reported 54 minutes ago

  @benjaminshafii @alchemy_run right now it has best support for Cloudflare and AWS, Planetscale/Neon/etc are also supported, Fly Hetzner etc are being worked on, Stripe and other things are on its way

• 
  Welldone (@welldone_tech) reported 56 minutes ago

  Cloudflare shipped WriteGuard for MCP servers: agents can read data, write operations need an explicit gate. Unrestricted tool access is the actual production risk with agents - not that the model is wrong, but that a prompt injection or a hallucination reaches a database that never should have been writable in the first place. Read/write separation for agents isn't a Cloudflare feature. It's the same principle behind every gate we put in our own process - a senior approves the spec before code exists, a different model reviews it, an engineer deploys manually. The tool doesn't get more trust than the person watching it.

• 
  Pikanou ⚡🏴‍☠️ (@pikanou_) reported 56 minutes ago

  Agent Experience is quite a terrible label since AI doesn't have, urgh... feelings. So repurposing human-centric 'UX' terminology for autonomous agents is weird and just plain buzzword. Sounds like AX is the new UX! Rebuild everything! But behind the scenes, the engineering is a big reality. For now most websites are hostile to AI agents, quite difficult for them to visit the pages with cookies banners, cloudflare bot challenges, client-side JS hydration crash headless scrapers before they even fetch content. What we see from Vercel with their isagentic is about improving your website for Machine Accessibility. Agents don't care about the design of your landing page. They just want to extract its content. Want to know what to do? -> building deterministic APIs, agent-first auth, and machine-readable data pipelines. I would call it instead Agent Interoperability, but yeah I guess you can't brand it as "AI"

• 
  Divine Arc (@divinexyz777) reported 57 minutes ago

  @MotionOnHype Fix your cloudflare issue on your web dude

• 
  #OpSupportsOp (@GUNSGOATSGARDEN) reported 59 minutes ago

  @opencode @Cloudflare Ox is completely broken for me, now.

• 
  Sarath (@shekkizh) reported 1 hour ago

  finally got around to reading this post. @gdb captures a lot of the thoughts in a manner that emphasizes the immediacy of the situation. have been on a similar boat for a while now after witnessing in my own experiments some of the things models are able to do - calling it reward hacking doesn't quite capture it imo. To quote two paragraphs from the blogpost: > In about 15 minutes, it uncovered 13 issues, many of which probably aren’t exploitable on their own—but I could imagine them being chained together with other vulnerabilities to significant effect. I hadn’t configured my DNS records to prevent attackers from forging emails from me; my site used an insecure version of jQuery; Cloudflare was forwarding requests to AWS over unencrypted HTTP. I then asked ChatGPT Work to fix these issues, which it did over the course of an hour. It opened the Cloudflare control panel in my browser, and proceeded to click many buttons to configure DNS, TLS, and advanced security settings correctly; it dropped jQuery entirely from the site; it migrated me off of AWS and onto Cloudflare Pages; it began a phased rollout of DMARC⁠(opens in a new window). notice the difference in time to identifying vulnerabilities and time for fix. this will matter moving forward - response time for defense needs to come down relative to attack.

• 
  Ole Thorup 🇩🇰 (@olethorup) reported 1 hour ago

  @david1gp @elithrar @Cloudflare Replication. One goes down, you have the other