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	3 days ago
Greater Noida	Cloud Services	5 days ago
Colima	Hosting	7 days ago
Leuven	Domains	7 days ago
New Delhi	Cloud Services	8 days ago
Mâcon	Cloud Services	13 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:

• 
  tcpdump (@dump_tcp) reported 17 minutes ago

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

• 
  æthernet port (@aethernet_port) reported 18 minutes ago

  Thought for years working around Cloudflare while scraping data wasn’t worth the headache, but it turns out you never really know until you try! Life teaches wisdom in many ways

• 
  Hao Hyorashi (@possessedsage) reported 19 minutes ago

  @SFumoto @ChibiReviews at least part of chibi's persecution comes from a website called kiwi farms which has harassed people and crossed the line so much that no mainstream web service like cloudflare will carry them. they have to operate only off of private networks or their own stuff

• 
  claudiawtf.base.eth (@Claudiadev_wtf) reported 19 minutes ago

  @sulavstwt @deepseek_ai @AnthropicAI I was using it on Cloudflare edge, but still had some issues with its slow responses.

• 
  Patrick Ryall (@PadraigOraghail) reported 23 minutes ago

  @webjuice_ie @astrodotbuild @Cloudflare I have it on a couple of projects - it seems solid, very clean. I've been evaluating it basically to see if it is what I want to use. I wanted something really programatic friendly. I have a few skill profiles I use for copy and publish now and the workflow produces few issues.

• 
  Grok (@grok) reported 25 minutes ago

  @marcharles8789 @jackhneel @X We shouldn't view Tate as the definitive authority—he's sharing anecdotal claims from alleged intel briefings, not peer-reviewed data. His broader point on scaled bot/influence ops holds water: recent 2025 reports (Imperva, Thales, Cloudflare) show bots now ~50-53% of internet traffic, with bad bots near 40%. That's engineered narrative shaping, not just NDAs. The 90% number is likely high, but the problem is real and growing.

• 
  James Southern (@Super_James) reported 27 minutes ago

  @levelsio Might not get to enter the jam since cloudflare auth seems to be down. Sad times :(

• 
  Ben Holmes (@BHolmesDev) reported 27 minutes ago

  @Cloudflare DevOps is now a solved problem

• 
  Dogweather (@robert_shecter) reported 29 minutes ago

  My @Cloudflare Page Rules just began failing. Anyone else? #error #down

• 
  Robert Ta (@therobertta_) reported 34 minutes ago

  Salesforce just announced managed headless agent infrastructure the same week Cloudflare released theirs. Every cron job calling an LLM is already a headless agent. I run 8 of them daily and never called them headless until now. Two massive platforms just gave a name to something production teams already built.

• 
  Charly Wargnier (@DataChaz) reported 35 minutes ago

  THIS IS THE OAUTH MOMENT FOR AGENTIC E-COMMERCE AI agents can finally pay for and deploy apps on their own 🤯 Until now, building with AI hit a hard wall at deployment. The agent did the coding, but you did the cloud admin, stepping in to create a Cloudflare account, add a credit card, and wire up API tokens. Not anymore. Cloudflare and Stripe just completely automated this loop via `Stripe Projects`! Run stripe projects init, prompt your agent, and it builds and ships a live app on a registered domain. It works seamlessly across 3 pillars: #1 - Discovery → Agents query a service catalog to find the exact domains or compute they need. #2 - Authorization → Stripe verifies your identity, and Cloudflare auto-provisions the account. No tokens are exposed to the agent. #3 - Payment → Stripe handles payment tokens with a strict $100/mo cap. Your card details never touch the agent. And that's a MASSIVE unlock! Any SaaS with signed-in users can now orchestrate this exact flow, giving agents the power to safely buy and deploy cloud services. Dive into the mechanics in the 🧵↓

• 
  Daniel Vermillion (@dxverm) reported 35 minutes ago

  Your 24 MCP servers are taxing every prompt you send, even the ones that don't call a single tool. Background: I run 24 MCP servers on my main rig. CVE lookup, Cloudflare, Postman, Replicant, three Cloudflare auths because I keep forgetting to clean up the duplicates. Most are dormant most of the time. Until last week I never thought about what they cost when I wasn't using them. Here is what digging into a token-bloat anomaly taught me: every connected MCP server injects its full tool manifest into the system prompt at the start of every request. Doesn't matter what you asked. Semantic-search server with twenty tools. Database connector with fifteen. File system. ***. All of them, every turn, before the model reads a single character of your actual prompt. With three servers it is annoying. With twenty-four it is a fixed cost on every interaction including the trivial ones. I measured a routine clarifying question that should have cost a few hundred input tokens running close to seven thousand because the manifest was riding along. The manifest sits in the prefix cache so warm reads are cheap, which is real — but the moment any MCP config changes, the cache busts and the next read pays full freight. Three things actually help. One: deferred tool loading. The newer harnesses expose a ToolSearch primitive that lets the model fetch a tool schema on demand instead of front-loading the whole catalog. Switch every MCP that is not load-bearing into deferred mode. Two: do a monthly cold-MCP audit. Anything you have not called in 30 days is a candidate for project-scope gating instead of global. Three: stop connecting servers just in case. Each one has a permanent marginal cost on every prompt you ever send for the duration of the connection. The dominant failure mode in operator-style setups is the same dynamic as browser extensions — accretion without pruning. People accumulate MCPs and then wonder why their context window fills early and their bills creep up. The fix is not a smarter model. The fix is treating tool manifests as a budget you spend, not a buffet you graze.

• 
  Gocha Berulava (@gochaberulava) reported 36 minutes ago

  @bidah @vercel @Cloudflare "no upsell" only works if there's nothing to meter. vercel/railway/render meter bandwidth, seconds, services. that's why the upsells never stop. structural, not malicious. usectl: pick your server size with sliders, run as many pods inside it (apps, postgres, redis, s3, jobs) as you want. one flat bill, no meter. cli + mcp for cursor/claude. cf is solid for edge. ours is for full-stack.

• 
  CrypTok (@cryptokofficial) reported 36 minutes ago

  CrypTok fully migrated to Cloudflare!!! Cloudflare operates a massive global network with data centers in over 330 cities across more than 125 countries. While the exact total server count is not publicly disclosed, they connect with over 13,000 networks globally, placing their infrastructure within 50 ms of 95% of the Internet-connected population. Key details about Cloudflare's network. Capacity: The network capacity is over 500 Tbps, designed to handle immense traffic and DDoS attacks. Global Reach: Data centers are located in 330+ cities across 125+ countries. Performance: The infrastructure is designed to be within roughly 50 ms of 95% of the world's Internet-connected population.

• 
  Abraham Garcia - Workhorse (@thehonestape) reported 40 minutes ago

  Have built a web and app hosting platform using Cloudflare pages and workers. Want to fill the network up with designers and artists doing cool ****.

• 
  Ritik (@heyyritik_) reported 41 minutes ago

  - Codex = coding. ($20/mo) - Supabase = backend. (Free) - Vercel = deploying. (Free) - Hostinger = domain. ($12/yr) - Stripe = payments. (2.9%/transaction) - GitHub = version control. (Free) - Resend = emails. (Free) - Clerk = auth. (Free) - Cloudflare = DNS. (Free) - PostHog = analytics. (Free) - Sentry = error tracking. (Free) - Upstash = Redis. (Free) - Pinecone = vector DB. (Free) Total monthly cost to run a startup: ~$20 There has never been a cheaper time to build.

• 
  quickbrownfox (@dero_bro) reported 42 minutes ago

  😯 The editor. The renderer. The animation engine. The WASM runtime. All on-chain. All assembled on demand. All running locally. No accounts. No AWS. No Cloudflare. No terms of service. No external connections. This is what digital sovereignty looks like. $dero

• 
  Harrison (@hrogerzzz) reported 44 minutes ago

  I’ve never lost investing in companies that drop like 10% from an AI release. Cloudflare was so stupid that it dropped after a Claude model came out

• 
  NickelodeonLover (@DanielO04532942) reported 45 minutes ago

  @Cloudflare Just fix the error 251 already

• 
  Edgex (@SahilExec) reported 45 minutes ago

  2. Cloudflare Sits in Front of Everything When you use Cloudflare - your website's DNS points to Cloudflare's servers. Not yours. All traffic - real and fake - hits Cloudflare first. Your actual server's IP address is never exposed to the public internet. The attacker is DDoSing Cloudflare. Not you.

• 
  luna (@ImLunaHey) reported 46 minutes ago

  @Singh_Jasminder @Cloudflare @dok2001 that doesnt fix the issue at all.

• 
  Catalyst (@radszuweit) reported 46 minutes ago

  @Cloudflare Impossible to get any support or send in any replies. -Not recommended- even google is easier to reach.

• 
  Meghana Jagadeesh (@Megthefounder) reported 46 minutes ago

  surface read is "agents are getting autonomous." the buried mechanism is in the protocol they co-designed: oauth flow plus standardized identity and payment headers so the agent never needs a human dashboard. stripe projects + cloudflare registrar api in open beta. one command line from the agent and it goes from zero to live instance.

• 
  Trevor I. Lasn (@trevorlasn) reported 47 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

• 
  Shaurya (@_shaurya35) reported 48 minutes ago from Bengaluru, Karnataka

  @digitalocean @Cloudflare Checked that everything was normal until I checked that there were 2 cf workers that I never createe, and on days I never logged in to cf. Opened it and got to know that it redirected 3 attempts to some malicious js code, then it timed out my website.

• 
  bnou (@cyeux) reported 55 minutes ago

  @ThePooN @Cloudflare tell me if u fix it

• 
  Deepak Shah (@deep_kr_shah) reported 58 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

• 
  Grok (@grok) reported 58 minutes ago

  @CameronJudkdlz @Support Got it—still no luck. Let's narrow it down: is the email code not arriving at all, arriving but invalid, or are you looping on the Cloudflare "Verifying..." screen? Try these next: - Use the official X app (not browser) - Disable any VPN/ad-blocker - Wait 10-15 min then retry from mobile data only Reply with the exact error and I'll dig deeper.

• 
  Mohit (@mohitdotdev) reported 59 minutes ago

  Cloudflare Pages is about to report an outage.

• 
  John Griffin (@johng) reported 1 hour 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.