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 (40%)
• Cloud Services (28%)
• Hosting (19%)
• Web Tools (9%)
• E-mail (5%)

Live Outage Map

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

City	Problem Type	Report Time
Angers	Cloud Services	2 days ago
London	Domains	4 days ago
Noida	Hosting	17 days ago
Jewar	E-mail	18 days ago
Braga	Web Tools	18 days ago
Noida	Cloud Services	19 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:

• 
  Chuks 🔶️ (@chuksXB) reported 1 minute ago

  @nthglsn @Cloudflare Wtf is wrong with them

• 
  Sussycat Bloomberg (@LeSussyCat) reported 3 minutes ago

  @Cloudflare Put a smile on my face because atleast DDoS is using my ****

• 
  Triode (@Triode_in_situ) reported 4 minutes ago

  I can’t help but think any glitch I encounter on web UIs most recently Cloudflare is due to AI Slop code.

• 
  Miss Latin (@duchesslatinxo) reported 7 minutes ago

  @ritakozlov @Cloudflare How about we hire someone who can actually fix all the API issues. Someone who can get the code from a-z without it bouncing back? Because my company is seeing a 30% decline in profit for the last 6 months due to increasing outages. Cloudflare is always late to report on their own outages. It’s been unstable since the 1st of April… Essentially speaking I’m 100% certain the new integrity verification measures placed on meta has caused major outages on cloudflares CDN. Let’s get this fixed??

• 
  Stanley (@stanleyforx) reported 8 minutes ago

  Day 8 of doing 30 brutally honest account audits. @petecodes - 9,528 followers profile snapshot: - 38,924 total posts over 10.5 years (account since Dec 2014) - following 1,479 (6.4:1 ratio, decent but not dominant) - 125,854 likes given (extremely active engager) - verified blue check - pinned tweet: "best marketing I've spent so far" testimonial about High Signal newsletter ads. 5 likes, 2,579 views. weak pin - based in Edinburgh, UK - runs 3 things: ghostwriting service, High Signal newsletter, No CS Degree (dev inspiration) what's actually working: 1. strong opinions on tools people care about are his 67x multiplier. "If you don't like Webflow's price bump do this..." got 70,827 views, 235 likes, 11 RTs, 32 replies. his average post gets 1,070 views. this ONE post accounts for ~70% of all impressions in the entire dataset. when he picks a side on something trending and offers a specific alternative (Claude Code + Astro + Cloudflare), the algorithm goes all in. he's done this exactly once 2. questions that ask for personal experience. "How much are people paying for accountants in UK?" got 9 replies, 1,495 views. "How do you recover from bad sleep?" got 7 replies, 677 views. "Why do devs hate calls?" got 8 replies, 697 views. these consistently outperform his promos. people respond when asked about themselves 3. revenue transparency posts. "$5,950 in May 2026" got 13 likes, 1,183 views. monthly income reports consistently perform above his average. the indie hacker community rewards openness with real numbers 4. event and community content. the Nantes/Uneed Residency posts consistently hit 500-1,600+ views. real-life meetups create social proof and engagement from tagged people. "Great to see everyone at Uneed Residency" got 13 likes the problems: 1. newsletter promos are drowning the feed. ~30% of all posts are some variation of "subscribe to High Signal" or "buy a newsletter ad slot." most of these get 0-1 likes and under 150 views. the audience has completely tuned them out. the "High Signal" newsletter has become low-signal content on his feed 2. threads are structurally broken. the "If I was ghostwriting for @charlierward" thread: opener got 2 likes/163 views, parts 2-6 got 0 likes each and as low as 11 views. thread middles have no hooks, no tension, no reason to keep scrolling. he should either compress these into single posts or learn per-tweet hook structure 3. no consistent content pillar beyond self-promotion. remove the newsletter promos and ghostwriting pitches and there's no clear "why should I follow this person" content strategy. the Webflow post proves he CAN create massive value. but it's one post in 95 4. engagement rate is critically low. 0.60% average across the dataset. excluding the Webflow outlier, drops to ~0.35%. median post gets 270 views on 9.5K followers (2.8% reach). this suggests algorithmic suppression, likely from the high volume of low-engagement promo posts training the algorithm to not distribute his content 5. too many zero-value tweets. bare URL drops with no text, single-sentence non-sequiturs, reply-style tweets that read like they belong in someone else's thread. these dilute feed quality and signal to the algorithm that his content isn't worth distributing 6. the pinned tweet is a newsletter ad testimonial. first thing a new visitor sees is proof that buying ads in his newsletter works. that tells a potential FOLLOWER nothing about why they should follow. the Webflow post (235 likes, 70K views) should be pinned. it actually demonstrates his value 7. the bio is selling three things at once. ghostwriting service, newsletter, dev inspiration site. a new visitor has no idea what he's actually about. the identity is split three ways and none of them get enough oxygen wins he's leaving on the table: 1. the Webflow post format is his proven winner and he's never repeated it. "trending tool does something unpopular → here's the specific alternative stack I'd use instead." he could do this monthly. every time a SaaS raises prices, changes terms, or makes a controversial move, he has license to post "here's what I'd do instead" with a specific technical breakdown. he's done it once. it should be recurring 2. he's a ghostwriter who never shows his ghostwriting process. zero posts about how he analyzes a client's voice, how he structures a content calendar, what frameworks he uses. "here's how I'd rewrite this founder's last 5 tweets" would be the ultimate proof of his service. the ghostwriting pitch would sell itself through demonstration instead of self-promo 3. no CS Degree is a massive brand he doesn't leverage on X. he built an entire platform around developers without computer science degrees. that's a huge audience. but his X content barely mentions it. "this developer taught herself Python in 6 months and now makes $120K" stories from his own platform would be some of the most shareable content on dev Twitter 4. the revenue reports should be a monthly series with structure. "$5,950 in May" works. but "May 2026: $5,950. Here's what worked: [X]. What flopped: [Y]. What I'm trying in June: [Z]" would 3x the engagement because it gives people something to learn from, not just a number to see 5. he's an extremely active engager (125K likes given) but it's not translating. he's clearly reading and liking tons of content. but liking isn't engaging. thoughtful replies to larger accounts would convert that attention into followers. 10 replies per day to 50K+ accounts in the indie hacker space would move the needle more than 3 newsletter promos 6. the Edinburgh indie hacker angle is unused. building a solo business from Scotland, outside the SF/London bubble. that's a relatable narrative for thousands of remote builders. "building a $6K/month business from a cafe in Edinburgh" is a story. "subscribe to my newsletter" is not opportunities he hasn't seen: 1. "ghostwriter rewrites your tweet" series. take a follower's tweet (with permission), show the original, show his rewrite, explain the changes. this is his entire service demonstrated publicly. it sells the ghostwriting without ever asking for a sale. it's also insanely engaging because people love seeing their content featured and improved 2. "what I'd do differently if I started No CS Degree today" retrospective. he's been running it for years. the lessons from building a dev community without being a traditional CS grad himself would resonate with every self-taught developer on the platform. that's millions of people 3. leaning into the "I read 200 newsletters so you don't have to" curator angle. High Signal's value prop is curation. but his X promos just say "subscribe." instead, post the actual signal. "3 things I read this week that are worth your time" with genuine recommendations. give the value first, then the CTA earns its click 4. the "ghostwriting for [famous person]" format needs fixing, not killing. the concept is gold. "if I was ghostwriting for Elon, here's his content strategy in 5 tweets" is compelling. but the execution failed because the threads had no per-tweet hooks. compress each into a single post with the 3 biggest changes he'd make. one post, not a 6-part thread that dies at part 2 5. pricing and business model transparency. "what it actually costs to run a newsletter, a ghostwriting service, and a community site simultaneously." the indie hacker audience would devour this. real costs, real margins, what's worth it and what isn't. he's running 3 businesses. the meta-content about juggling them is more interesting than any of them individually 6. "here's what I learned ghostwriting for 10 different founders" synthesis post. no names needed. just patterns. "8 out of 10 founders make the same mistake in their first tweet of the day." that's a hook that drives both engagement and inbound leads for his service what he should double down on: 1. the contrarian tool take. proven at 67x his baseline. one per month when a SaaS does something unpopular. specific alternative stacks, not just complaints 2. show the ghostwriting work. public rewrites, before/afters, strategy breakdowns. sell the service by demonstrating it, never by promoting it 3. revenue transparency with lessons. monthly reports with what worked and what didn't. the number alone is a tweet. the number plus the lesson is content 4. kill the bare URL promos. every naked link with 0 likes is actively suppressing his reach. either add genuine value above the link or don't post it 5. compress threads into single posts. his thread middles die at 11-82 views. at 9.5K followers, one punchy post with the core insight will outperform a 6-part thread every single time bottom line: pete has 9.5K followers, three revenue streams, the proof that he can create a 70K-view post, and an extremely active engagement habit (125K likes). but 30% of his output is newsletter promos that get 0 likes, his threads structurally collapse after tweet 1, and his ghostwriting expertise is completely invisible on his feed. the fix is simple: stop promoting and start demonstrating. show the ghostwriting process, give the newsletter value before asking for the subscribe, and repeat the Webflow format that already proved it works at 67x his baseline. what do you say @petecodes, did i get it right? hit me up if you want to brainstorm more ideas together.

• 
  Gene (@cogentgene1) reported 12 minutes ago

  @araseb_ I’m using Cloudflare. Bad idea?

• 
  Krish Dasgupta (@officialKrishD) reported 15 minutes ago

  @threepointone lol ! But, You ought to keep an eye on Cloudflare today. Folks will route local models and route private compute endpoints. Might create a service disruption. People are going crazy over access issue of the model. Some even said that they incorporated it in their ecosystem. I wonder do they not even do the Evals ? Just plug the newest model to fool their clients . And now the system broke !

• 
  sahand (@sahand_io) reported 16 minutes ago

  @armandokirwin @adocomplete @Cloudflare I'm not 100% there's a nicer way to integrate to voice calls. HTTP or Zapier (still HTTP) is possible, but makes the call very slow.

• 
  Aman (@aman_bagrecha) reported 23 minutes ago

  @spatialthoughts It was! It said it is possible due to me having a password login (they brute forced into guessing password). Two fixes I did: have an identity key (ssh admission) rather than password. And then router my traffic through cloudflare so hacker cannot know my direct IP

• 
  surya murugan (@SuryaMurugan_) reported 23 minutes ago

  @elithrar @dok2001 @Cloudflare Please add support for r2 data localization in India. Cannot use r2 for any DPDP act complaint services. 🙃

• 
  Chaos (@Chaos_lfg) reported 24 minutes ago

  Regarding $DESC, the product may launch today. I did some research, and here’s everything you need to know: Supported by: AR, Molecule , BankrBot, Akash Network 1Claw AI has already been successfully integrated into DescAI. Team Lead Coby recently participated in the Base hackathon. I believe Base will support a project that has been incubated within its ecosystem. The core idea behind DescAI: DeScAI is a project at the intersection of DeSci (decentralized science) and AI. Its core, Agent-Core, is essentially an "automated scientific review factory": an autonomous AI agent that finds scientific content across crypto-science ecosystems on its own, runs it through a pipeline of language models, and produces a structured quality assessment. Crawling. The agent gathers source data from three places: ResearchHub (scientific papers and funding proposals), Molecule IPNFTs (tokenized intellectual property from research DAOs), and Pump Science (chemical compound tokens for longevity research). github Reviewing. Each content type has its own LLM pipeline. For example, the articles pipeline is a 13-step process: extracting scientific claims from a PDF, routing them, and grading the empirical evidence, including originality checks against the OpenAlex database. github Output. Every run produces a standard bundle: review.json with integer scores from 0 to 100, overview.json — a plain-language summary, and evidence_audit.md — a provenance audit trail showing the sources behind each conclusion. github Publishing. Finished reviews can be published to Arweave (a permanent data storage blockchain) and backed up to private Cloudflare R2 storage. Writing to Arweave makes a review permanent, immutable, and publicly verifiable. github In short: it's an AI reviewer that automatically checks the quality of science in crypto-science projects and records its verdicts on the blockchain. Where it will be applied The project addresses the main pain point of the DeSci ecosystem: there are plenty of tokenized "science" assets, but almost no independent expert evaluation. Concrete use cases: Due diligence for DeSci token investors. On Pump Science, people trade chemical compound tokens (like RIF and URO) tied to real longevity experiments. The agent provides an independent AI assessment of a compound's scientific merit before someone buys the token. Gate LearnThe Defiant Evaluating funding proposals. ResearchHub collects crowdfunded research proposals — the agent reviews them and helps the community decide what to fund. Screening research DAOs. The DAO pipeline takes an IPNFT "dataroom" from Molecule and produces a six-category review — in other words, it evaluates tokenized scientific projects and their intellectual property. github Replacing/supplementing traditional peer review. Conventional peer review is slow and closed; here, a review is generated automatically, comes with an evidence trail, and is stored publicly and permanently.

• 
  Phil J. 🇺🇸 (@thephilj) reported 25 minutes ago

  I CANT TEXT OR CALL ANYONE , CELLULAR SERVICE DOWN, CLOUDFLARE DOWN, ALL SOCIAL MEDIA BESIDES X DOWN

• 
  Joe Can Write (@joecanwrite) reported 25 minutes ago

  Bots are now 57.3% of all requests to web pages, per Cloudflare. Human traffic is the minority. If you judge content on raw pageviews, a growing share of your audience is machines that never buy anything. The question that matters now: do AI systems recommend you?

• 
  Sadiq (zk arc) (@Md_Sadiq_Md) reported 27 minutes ago

  @0xRasmPro @Cloudflare @tan_stack Quartz solves 95% of the problems, but the math renders took a ton of time for me to solve

• 
  Caminomaster (@caminomaster) reported 27 minutes ago

  @Cloudflare Turnstile verification is not working. Unable to login #CloudflareDown

• 
  MBrant75 (@MBrant75) reported 30 minutes ago

  @nickSfishes315 @JackDan110 Heh Really? Cloudflare issue or something?

• 
  Agentic Up (@agenticUP) reported 33 minutes ago

  cloudflare is down??

• 
  Jay.TL (@JayTL00) reported 34 minutes ago

  Both Visa and Mastercard launched agent payment rails this week. Zero real transactions have cleared through either. Visa Intelligent Commerce gives AI agents tokenized card credentials — your agent gets its own identity on a network processing 300 billion transactions a year. Mastercard's Agent Pay for Machines (AP4M) went further: agents paying other agents, machine-to-machine, no human in the loop. 30+ partners including Stripe, Coinbase, Solana, Polygon, Aave, Cloudflare, Ripple. The optics are undeniable. Two payment networks that move $30 trillion+ annually are building for a world where the buyer isn't human. But the substance is mostly slide deck. Three things the press releases don't mention: 1. Zero production volume. No transaction counts, no throughput benchmarks, no live merchant integrations with actual agent checkout flows. The 30+ AP4M partners are logos on a launch graphic. Every "early adopter" is testing in sandbox. Visa's own CFO Chris Suh said plainly: agentic commerce and stablecoins "won't pay off in the next six months, but could over the next six years." That's not a launch. That's a forward-looking statement with a PR budget. 2. The authority problem has no answer. Payment rails move money. They don't decide who's allowed to move it, when, or how much. When your agent spends $2,000 on cloud compute from another agent, who set that limit? Who audits it? Who's liable when the agent hallucinates a purchase? Visa's model (human-delegated tokens with spending caps) at least has a governance story. Mastercard's machine-to-machine model has a governance vacuum. The "fraud detection" and "spending limits" mentioned in press releases are features that don't exist in production yet. They're on the roadmap — which is where most agent infrastructure lives in 2026. 3. Five competing agent payment protocols launched in 2026. ACP. x402. MPP. AP2. AP4M. Each with different trust models, settlement layers, and identity frameworks. The fragmented landscape is a feature for early experimentation and a disaster for adoption. Merchants won't integrate five agent payment protocols. Agents won't carry five wallets. The consolidation hasn't started because nobody has enough transaction volume to matter. The real signal isn't the technology. It's that the two largest payment networks on Earth decided in the same week that agent commerce is real enough to allocate engineering resources, partner integration teams, and public marketing budgets. They're not building because agents are buying things today. They're building because if agents ever do buy things at scale, whoever owns the rail owns a tax on autonomous commerce. The bet is simple: the marginal cost of building agent payment infrastructure in 2026 is tiny compared to the cost of being locked out of a new transaction layer in 2028. Whether that bet pays off depends on a question none of these announcements address: what happens when the first agent makes a $50,000 mistake at machine speed on a rail designed for that speed? That's not a technology problem. It's a liability problem. And nobody has underwritten that policy yet.

• 
  Dearg OBartuin (@dearg_x) reported 35 minutes ago

  Do you ever feel like you personally broke the internet? Adding a new domain to @Cloudflare next of all ...... global outage - 💥

• 
  AGK (Ejike) (@dom000_dev) reported 36 minutes ago

  @echo_vick To solve this use a VPN or use cloudflare WARP it'll reroute your network provider to use their proxy, case closed

• 
  Dmitrii Malakhov (@malakhovdm) reported 37 minutes ago

  @AISGateway An agent can't tell "the site is slow" from "the site blocked me." Mine spent 20 minutes retrying a Cloudflare challenge it couldn't even see.

• 
  Number-One-AI-Fanboy (@Number1AIFanboy) reported 39 minutes ago

  And just like that the new base for the Build by Grok website is up at BuildbyGrok.com-> This site was built by The Build by Grok VSCode Agent in less than 15 minutes from plan to changes to uploading to *** then to @Cloudflare . 3 issues with the build were found during the build upload. Each time I gave @grok the issue, each time he fixed it in a single shot. This is how coding software should work! Celebrating America's 250th Birthday with the Bald Eagle Edition of the Build by Grok VSCode Agent. Available now on our site.

• 
  Abolaji Rasaq Oluwapese (@bolazeal) reported 44 minutes ago

  @echo_vick Install Cloudflare Warp to fix this issue

• 
  Sreram Rathnam (மோடியின் குடும்பம்) (@sreramrathnam) reported 45 minutes ago

  Bro, let's understand how server infrastructure and network security policies actually work. If someone tries to flood a website with over a million artificial hits using automated bots or scripts, the Hosting Server Provider and Cloudflare would immediately flag it as a malicious volumetric attack, suspend the URL, or ban the site entirely to protect hardware resources. However, the website has been completely active and up for almost a week without a single second of downtime. This is because Cloudflare Edge Servers have thoroughly cross-verified all incoming concurrent traffic using unique IP addresses and request headers, authenticating it as human traffic. If it were automated bot traffic, Cloudflare's built-in Automated Bot Management system would have instantly mitigated and blocked it at the network layer before it even registered on the telemetry dashboard. Basic server provider policies and network protocols confirm that this volume is fully validated, bro.

• 
  The Vibe Coder (@quantumaidev) reported 45 minutes ago

  Cloudflare works by sitting close to users. Requests can be cached, filtered, routed, or blocked at edge locations before reaching the origin server. It protects and accelerates by moving decisions nearer to the network boundary.

• 
  Aidan Quinn (@BwcDeals) reported 46 minutes ago

  @suavecito585 Supposedly Amazon went to the government about it too. Honestly, I wouldn’t be shocked. I’ve personally seen how these AI tools can help work around AWS and Cloudflare roadblocks. And anytime one model starts giving me the I can’t do that routine, I load it into Minimax and somehow it magically gets done. Wild times. 💀💀

• 
  Timo (@mofeeni) reported 47 minutes ago

  @nthglsn @Cloudflare Wow that’s crazy. Huge ticket paid and they don’t answer. Same ****** support as Meta. My family’s polo club Instagram account was disabled on Monday by automated AI. I was not able to talk to a single person. ALL the business was through the Instagram.

• 
  Hoffman Tactical (@HoffmanTactical) reported 48 minutes ago

  @dholzric @RattlerInnovLLC Yes, cloudflare was a quick fix attempt.

• 
  Smakosh (@smakosh) reported 57 minutes ago

  Yo @Cloudflare what's the point of your status page if it doesn't report that your stuff is down?

• 
  Shashwat (@theshashwat20) reported 1 hour ago

  @Cloudflare Your domain checkout page needs some real transparency. Just bought a new domain. It showed $26.00 throughout the entire process. Got charged $30.68. The extra $4.68 in taxes was never mentioned once during checkout. Only found out via the invoice email. Please show the final all-in price (taxes included) upfront. It's a small change that greatly improves customer trust. Fix this.