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 (36%)
• Cloud Services (31%)
• Hosting (17%)
• Web Tools (11%)
• E-mail (6%)

Live Outage Map

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

City	Problem Type	Report Time
Manchester	Domains	2 days ago
Angers	Cloud Services	14 days ago
London	Domains	16 days ago
Noida	Hosting	29 days ago
Jewar	E-mail	29 days ago
Braga	Web Tools	29 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:

• 
  bigdatachads (@bigdatachads) reported 2 minutes ago

  I've been building AI phone agents on @Cloudflare for a while now. v1. a Python container, fighting for every millisecond. v2. no container, the whole call on the edge. that was the real work. now that I have the stack down, I spent last weekend messing around. this is v3, a cartoon you talk to that remembers you and gets heckled by a second AI. all on Cloudflare primitives. three teardowns, first one tomorrow. follow along. @CloudflareDev

• 
  pathik (@pathikghugare) reported 8 minutes ago

  @NotRoodraksh @4k_isn not working on cloudflare warp

• 
  Errrg-XL (@ergonautgg) reported 9 minutes ago

  @TapTools clearly has zero plans to open source the services they brought to Cardano. What exactly was their problem keeping this site afloat (dont you use Cloudflare?) after receiving hundreds of thousands in treasury/Catalyst funding?

• 
  SpikeViper (@spikeviper) reported 10 minutes ago

  I am once again asking @Cloudflare for a response on why their support is radio silent on what is now a shady billing situation

• 
  Coffee and Gun Oil (@coffee_oil) reported 10 minutes ago

  @ShamashAran I was on cloudflare ******* with DNS last night. I hate DNS The **** I run locally works fine, but that's because it's me and a text file.

• 
  sw1tch.sh (@thinkistillcare) reported 12 minutes ago

  suddenly i start getting captchas on google and cloudflare it has to be my IPTV service on my fire stick right?

• 
  zeb (@zebassembly) reported 15 minutes ago

  @astuyve @boristane Not to get too into the weeds but the concern is where the trace context gets inherited and where we check the users tracing configuration. Before a request ever goes to the Workers runtime there's our FL2 (essentially the Cloudflare webserver) that actually accepts the http connection for various reasons we want that part that isn't entirely related to Workers to be aware of tracing so we can do cool things in the future. This entails creating a way for FL2 to fetch the user's tracing config (sampling, if they want to enable propogation, etc), passing that context through FL2, passing it to the Workers runtime, and then when the Worker does a fetch we need to pass it back through to FL2 so we can potentially attach the context header. None of this is strictly about just parsing that trace context header, more about threading configuration and ceoss-service communication.

• 
  Patrick St. Jean (@stjeanp) reported 15 minutes ago

  @EddCoates I dealt with some of the same stuff, ended up putting them behind Cloudflare proxies, which helped somewhat. The biggest fix was blocking specific ASNs. Specifically AS132203.

• 
  FreeSolDev (@FreeSolGamesDev) reported 15 minutes ago

  We do a little upgrading ProofNetwork already includes native support for Cloudflare R2 as a Smart Service, but with network activity continuing to accelerate, I decided to expand its role within the infrastructure Transaction archives now leverage additional R2 backed redundancy, providing a stronger foundation as we push toward what I expect will be well beyond 10 million transactions Also spent time building an automated contract validation pipeline that acts as a pre deployment safety net Deployments are now automatically checked for common security oversights and best practice violations, with contracts being blocked from deployment if critical safeguards are missing, such as signature verification on privileged admin functions and other potentially exploitable patterns.

• 
  Caleb Sylvest (@calebsylvest) reported 17 minutes ago

  @jasondoesstuff Skip the CMS. Recently did the same. Used Claude to build everything Used Astro. Deployed to Cloudflare. Writing with MDX. Pre-rendered everything and served from Cloudflare edge network. Basically a fast as possible.

• 
  kocer (@kocer_eth) reported 18 minutes ago

  7 FREE AI API/TOOL TIERS YOU CAN USE TODAY BEFORE BUYING ANOTHER AI SUBSCRIPTION If you build agents, bots, research tools or small automations, start with this stack. 1. OpenRouter Use it as the router. It exposes free-priced models in the model list, so you can test routing before paying per token. 2. Google AI Studio / Gemini API Good for prototypes, evals, long-context tests, and agent experiments. Check the free tier before you burn paid credits elsewhere. 3. Cloudflare Workers AI Best when you want inference close to your app. The useful part is not just “free AI” — it sits inside the same place you can deploy Workers. 4. GroqCloud Use it when speed matters. Great for bots, voice loops, extraction, and any workflow where slow responses kill the demo. 5. GitHub Models Best for prototyping inside the GitHub flow. If your code, prompts, and tests already live there, this removes friction. 6. Tavily Research/search API for agents. Free plan shows 1,000 API credits/month, useful for browsing agents and research bots. 7. ElevenLabs Voice layer. Free plan shows 10k credits/month, enough to test narration, agents with voice, and demo content. > My rule: never build production on a free tier first. > Use free access to test: - latency - rate limits - output quality - tool calling fit - billing behavior - whether your agent actually needs the premium model Then pay only for the part that survives real usage. Most people skip this and buy 3 subscriptions before they even know which API call matters.

• 
  That Boosted Snep 🔜 Megaplex (@witch_the_snep) reported 22 minutes ago

  @joesmith1457 @DoorDash_Help It looks like its cloudflare thats having issue. That who hosts doordash’s website

• 
  POKE6900 (@poke6900gg) reported 25 minutes ago

  We are aware the website is down and due to this the apps aren't working as they should. This is due to a Cloudflare issue and we are working on a solution to get everything back online a.s.a.p.

• 
  Zerops (@zeropsio) reported 26 minutes ago

  @shubh19 @isha_singh06 Hey! Zerops fits the Railway/Render slot when the backend needs managed Postgres/Valkey on a private network next to it, hardware-priced. Pair it with Cloudflare for static the same way.

• 
  David Herrmann (@herrmanndigital) reported 36 minutes ago

  Meta ads in 2015: 1. Take a photo of product with iPhone. Open Power Editor, launch ad in under 2 mins. 2. Refresh PE and see 1:1 conversions in under an hour, scale within 3-4 hrs. 3. Spend the rest of the day coming up with fun ideas for new offers, new landing page ideas, new product ideas, new copy ideas. Meta ads in 2026: 1. Hire creative strategist -> Who hire creative agencies / editors -> who build batches of creative. 2. Open ads manager, go through 25-30 prompts of things you have really no clue about, but Meta is insistent it'll lower your CPA by 7.9% based on a study from 2 advertisers in 2022. Click publish after 1 hr of building out 10 new concepts. 3. You're then hit with an error publish with messages like: "you have no shops select," or "you don't have page permissions," or "this creative has expired, please try a new one" you finally are able to hit publish. 4. Finally, you can go on with your day. But wait, client sees that they are getting discounts on the ads you just launched coming in. You open up ads manager and go to ad level to go through multiple areas to figure out why. 5. Ah, Meta has randomly turned the promotions automatically on. No worries, you pause it. On with your day. 6. Someone who knows someone at the company for whose ads you just launched flags to friend that the ads are appearing with a dubstep version of a Beethoven song and seem out of touch for the brand. You're then texted to go pause the music. 7. Open ads manager, navigate to ad level, pause music. You're then hit with a prompt from Meta: "WHY ARE YOU TURNING THIS OFF?!" You hastily respond, "BECAUSE IT SUCKS!" 8. Visibly irritated now over Meta you finally settle in to do the other part of your job, analysis! 9. But wait, there appears to be an outage! Everyone texting, "Is Shopify down?," "Is it Cloudflare?," "Is it AWS?," "No! It's Meta!" Ok, well now you're in fight of flight. Should we wait? Should we pause? Let's wait, yeah let's wait. This is just a minor blip. 10. Meanwhile you're not left refreshing things every 5 min instead of doing analysis and research. Finally, it is night time and you can relax (not)! On this particular day you managed to launch 10 ad sets and do nothing else. Feeling defeated because what once took 2-5 min to do now feels daunting. It feels like you now spend the majority of your day in defense mode. Not because the work has changed or the people, but the process for doing it all has. And that's why you're feeling burnt out media buying friends and creative friends.

• 
  Flandermaxx (@Flandermaxx) reported 37 minutes ago

  A 28 year old Chinese engineer in Singapore bills 11 SaaS startups $63,400 a month for inference they think is running on AWS H100s. Forty NVIDIA Jetson Orin Nano boards stacked inside three IKEA boxes on the floor. A used Quest 3 on the bed. A half empty can of Yeo's chrysanthemum tea on the windowsill, still cold. The whole farm draws less power than his electric kettle. Each Jetson runs Llama 3.3 70B through MLC LLM, quantized to 4 bit. Each one serves embeddings, classification, and draft outputs at $0.40 per million tokens. OpenAI charges $0.60 for the same. A Cloudflare worker rewrites the response headers to read like an AWS us-east-1 region. The startups never check. They never asked. pause at 0:22, the camera holds on the IKEA boxes for two seconds. Everyone saw moving boxes. Almost nobody saw the holes drilled in the back. They were not for shelf pegs. They were the GPU intake. $63,400 in. $0 OpenAI bill. Hardware paid for itself in 11 weeks. His dad still thinks he is studying for the GMAT. He still calls home every Sunday at 8 p.m. He still says he has not picked a school. He still wears the same Uniqlo hoodie in every video call. He still has not mentioned the Stripe dashboard. A data center has racks, cooling, redundancy. He has three IKEA boxes, a kettle, forty boards humming quieter than the AC.

• 
  Brendan Irvine-Broque (@irvinebroque) reported 37 minutes ago

  dex is holding his Cloudflare product feedback hostage until I share this video watch so that I can learn what bugs we should fix

• 
  Hunter Bertoson (@BertosonHunter) reported 37 minutes ago

  @jamesqquick Watched the network tab, reverse-engineered an undocumented API, and turned it into a Cloudflare Worker that catches failed attendance syncs and emails an alert every night. Workers + cron is unreasonably good for this kind of thing.

• 
  Christopher Johnson (@cj_enlighten) reported 38 minutes ago

  Vanilla web search in an always-on agent gets blocked. Not a Hermes bug. A structural 2026 problem. Cloudflare and Akamai are aggressive enough now that any general-purpose agent hits the wall. You need Tavily or equivalent. Budget time for it.

• 
  Julian Goldie SEO (@JulianGoldieSEO) reported 44 minutes ago

  HERMES AGENT JUST TURNED WHATSAPP INTO A REAL AI WORKER But one tiny setup mistake can leave the entire bot completely silent. What Just Dropped: → Official WhatsApp Business Cloud integration → Run Hermes as a private assistant, team bot, or customer support agent → Text one WhatsApp number and get AI replies directly inside the chat Built Like A Proper App: ✓ Secure webhooks and official Meta support ✓ Voice notes, media, read receipts, and typing indicators ✓ Interactive approval buttons when Hermes needs confirmation The Setup: → Run `hermes whatsapp cloud` → Connect your Meta Business account and WhatsApp number → Use Cloudflare Tunnel to expose your local Hermes gateway securely Three Mistakes To Avoid: ✓ Use the 15–17 digit Phone Number ID, not your real phone number ✓ Replace the 24-hour test token with a permanent system-user token ✓ Subscribe to the `messages` webhook field or the bot will never receive anything For a reliable WhatsApp agent, use the official Cloud API path instead of the unofficial personal-account bridge.

• 
  Luca Steeb (@steebchen) reported 44 minutes ago

  @baanish @fayazara it's actually not true, you can use the CloudFlare AI gateway by setting it up in the dashboard and you'll get an URL which works with any SDK or library. however, personally I recommend to use @llmgateway as we support the full catalog of models and DevPass coding plan for 3x usage

• 
  Maaz Perwez (@MaazMz) reported 46 minutes ago

  @Aurarri How is it easier to install another app and then turn it on rather than doing it inside the app for which I want to use proxy? Plus cloudflare will control all of my network while telegram proxy only changes telegram...

• 
  Kamil Fabian (@KamilFabian) reported 46 minutes ago

  @EddCoates cloudflare free tier. Solved my 1mil req per minute problem. Just make sure to set up properly.

• 
  Nav Toor (@heynavtoor) reported 46 minutes ago

  The Dead Internet Theory was a conspiracy. The idea that the internet is no longer human. That bots and AI have quietly replaced real people. It started on anonymous message boards in 2019. Most people dismissed it. Stanford, Imperial College London, and the Internet Archive just measured it. They used the Wayback Machine to scan every new website published between 2022 and 2025. Thirty-three months of the internet, captured and classified. They applied one of the most advanced AI text detectors in the world to every page. 35.3% of all newly published websites were AI-generated or AI-assisted. 17.6% were completely AI-generated. No human involvement at all. In late 2022, before ChatGPT launched, that number was zero. In three years, more than a third of the new internet became synthetic. Not over decades. Not over a generation. Three years. Then they measured what that is doing to the internet itself. Semantic diversity is falling. The range of ideas, perspectives, and ways of saying things is narrowing. As AI content increases, the internet sounds more and more like one voice. Because it is one voice. The same models producing the same patterns across millions of pages. Positive sentiment is rising. Everything sounds upbeat. Polished. Confident. Helpful. The internet is getting friendlier while getting emptier. The tone improves as the substance disappears. The lead researcher, Jonáš Doležal at Imperial College London, said this to 404 Media: "I find the sheer speed of the AI takeover of the web quite staggering. After decades of humans shaping it, a significant portion of the internet has become defined by AI in just three years." Separately, Cloudflare reported that nearly a third of all internet traffic now comes from bots. Imperva reported that automated traffic surpassed human traffic for the first time in 2024. If you read my previous threads on Model Collapse and Retrieval Collapse, this is the final chapter. Model Collapse showed that AI trained on AI gets dumber. Retrieval Collapse showed that search engines indexing AI content get emptier. This paper shows the source of both problems. The internet itself is being replaced. The researchers are now working with the Internet Archive to build a live monitoring tool. A real-time tracker of how much of the internet is human and how much is not. The fact that we need a tool to measure how much of the internet is still real is the finding.

• 
  sunil pai (@threepointone) reported 48 minutes ago

  is there interest in a 4k+ word deep dive in building reliable agent loops (on cloudflare and elsewhere) writing down what I've done for building agents resilient to catastrophic failures on clients/servers/inference (with zero user code) and I need to get it out of my brain

• 
  Tushar Dwivedi (@tush_2708) reported 53 minutes ago

  @SayantikaSays Try @BSNLCorporate . And I am not joking. It is best if it's available in your area. No BS service. I never heard these lines from BSNL, which I kept hearing from other providers like Jio and Airtel. "We won't give you the router password" "You can't use your own router, even if ours is ******. What do you mean you have a better router and modem? Hou must pay us a subscription for this mesh thing instead" "No, you can't use Google or CloudFlare's DNS, you must use ours. What if it's slow? If you change, how will we snoop on you and inject our advertisements into your browser?" "No, you can't open a port for you. For that, you must buy a static IP from us. Why is a static IP needed to open a port? Who knows. But we won't allow it unless you pay us extra" In the last 3 years, I have only had 2 outages on BSNL, 30-40 minutes max. While on Airtel, I had caught their staff removing my cable from their box and adding a new one for a new connection, and then making me wait till they finally got a new box after a week. They just didn't want to make a new customer wait, so they simply assigned my slot to them. And their customer care and local staff wasn't even ready to accept it, unless I showed the CCTV footage and a video of my cable literally being thrown on the side of their box, not even connected, while they were claiming that there's some backend issue.

• 
  NOVA 🇷🇺 (@NOVA360HD) reported 54 minutes ago

  📌 The Illusion of Decentralization: Who Owns the Backbone of the Internet & AI in 2026? (Updated List) As you scroll daily, you might think you're navigating thousands of independent sites and apps. The reality? 90% of global data traffic flows through channels controlled by a select few. Here is who actually controls the world's digital backend: 1. The Cloud Big Three If these three companies went offline, half of the internet’s apps, banking systems, and aviation networks would vanish in seconds: * Amazon Web Services (AWS): Controls roughly a third of the entire global cloud market. It hosts giants like Netflix, Airbnb, and even highly sensitive government databases. * Microsoft Azure: The largest backbone for massive corporations, government institutions, and global digital identity systems. * Google Cloud: The third engine powering YouTube, massive big data research, and global startups. 2. The Gatekeepers These are the invisible shields you rarely see, but they control and protect your access to the internet: * Cloudflare: Manages and secures roughly 20-25% of all global web traffic. If Cloudflare goes down, half of the world's news outlets and crypto exchanges drop with it. * Akamai: The oldest and largest Content Delivery Network (CDN) in the world. They dictate how videos, live streams, and games reach billions of people without lagging. 3. The Hardware Monopoly Software is useless without processors, and this is where the greatest monopoly lies: * NVIDIA: Controls over 80% of the AI chip and data center GPU market. They essentially decide who has the compute power to train AI (like OpenAI and Meta) and who gets left behind. * TSMC (Taiwan Semiconductor): The Taiwanese giant that manufactures almost all the world's advanced chips for Apple, Nvidia, AMD, and Qualcomm. If TSMC stopped, the production of global smartphones and military hardware would freeze. 4. The Submarine Cables (Who Owns the Physical Internet?) The internet isn't in the sky (satellites only cover a tiny fraction). 99% of global data travels through cables at the bottom of the ocean: * SubCom & ASN: The two companies responsible for laying and maintaining most of the world's underwater fiber-optic cables. * The Big Tech Alliance: Today, Google, Meta, Microsoft, and Amazon have become the largest investors and tenants of these cables, meaning they now physically own the routes data takes between continents. 5. The Institutional Masters If you dig deep into the shares of every media, tech, aviation, and defense company globally, you will always find three names repeating as the top institutional investors: * BlackRock (Manages over $10 Trillion in assets). * Vanguard Group (Manages roughly $8 Trillion in assets). * State Street These funds don't run the companies day-to-day, but they hold massive voting power to dictate board members and the strategic direction of these giants (from Apple and Microsoft to oil and defense contractors). 💡 The Bottom Line: The internet is no longer the free, distributed network it was once touted to be. It has evolved into a highly centralized infrastructure where a few massive corporations and investment funds dictate what you see, what you hear, and how your data flows.

• 
  special k | CEO of stressed out era (@specialkdelslay) reported 56 minutes ago

  @HeadmasterDuck We have cloudflare pro acc which does mitigate some of this. The cloudflare dependency everyone has is a problem tho

• 
  starmex (@starmexxx) reported 58 minutes ago

  WHY WASTE 16 MINUTES OF YOUR TIME ON THIS AI ENGINEER EUROPE TALK WHEN I CUT THE 5 BEST MOMENTS INTO 4 MINUTES FOR YOU bright data engineer exposed why your ai agent lies about searching the web. cloudflare blocks 20% of web from ai. 60% of chatgpt citations are broken. agents hallucinate instead of saying "i can't" 00:00 - llms are programmed to please. they make things up instead of saying "i can't" 00:42 - cloudflare blocks 20% of web. 60% of chatgpt citations are broken 02:03 - gpt-5 fails all 5 web tasks without proper tools. zero out of five 02:42 - cloudflare labyrinth feeds ai fake data. bigger hallucinations 03:13 - don't parse with llm. build a parser. saves 99% of tokens bookmark this and watch the supercut below

• 
  Ciccio Diddo 🐧🇪🇺🚽 (@CiccioDiddo) reported 1 hour ago

  @EddCoates another interesting cloudflare function is AI labirynt ... it create a ton of fake content filled with nofollow links generated by some **** IA... so the scraper spends a ton of resources to traverse all links then, if not enough start to ban by AS number instead of single IP