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	21 hours ago
New York City	Cloud Services	20 days ago
Los Angeles	Cloud Services	21 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:

• 
  Felix (@Reezxy23) reported 4 minutes ago

  There has never been a cheaper time to build. Claude = coding. ($20/mo) Supabase = backend. (Free) Vercel= deploying. (Free) Namecheap = 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

• 
  Welldone (@welldone_tech) reported 6 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.

• 
  Sarath (@shekkizh) reported 7 minutes 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.

• 
  shydev (@shydev69) reported 7 minutes ago

  biggest tech bubble that has burst in the last 6 months - cloudflare has poor ui/ux compared to vercel

• 
  Michael Flux (@michaelflux) reported 10 minutes ago

  @levelsio My own secret sauce is a tiny utility I built; - Every 15 minutes a Worker hits the Cloudflare Observability API, checks for any new error-level events since the last poll, and surfaces them (with traceId, service, message, errorId, etc.). - I get a clean list of “here’s what broke in the last 15 minutes.” I fix it. Done. No noise. No dashboards. Just signal.

• 
  Anshu (@101xanshu) reported 11 minutes ago

  @MattieTK @Cloudflare yeah i just kept finding workarounds instead of opening issues lol. should've filed them properly, will do next time

• 
  Jon ONeill (@HouseHackerJon) reported 12 minutes ago

  On the software company side an issue I’m seeing building with @Cloudflare is that as a start up everything is aimed at @vercel and @supabase, but I’m a big fan of having everything run through Cloudflare once I discovered its system

• 
  Southpaw | ZZZ Optimizer v6 is LIVE (@Southofpaw) reported 12 minutes ago

  @JaIdabaoth Nevermind sorry just checked enka and it’s up. I might need to check cloudflare to see there’s an issue

• 
  TheDataBunny (Darth Bunny) (@thedatabunny) reported 17 minutes ago

  What would you do if cloudflare and AWS was down for a week?

• 
  developing valhalla (@valhalla_dev) reported 18 minutes ago

  Notably, most technology that has been lifechanging and revolutionary is technology that the average person does not at all care about. The average person doesn't care about Linux, or Docker, or the HTTP protocol, or virtual hosting, or React. They don't have to. They like what comes downstream of that technological progress. So the argument that "AI doesn't matter to most people" is moot, because the downstream effects of AI do matter. Yes, AI People need to stop pretending like AI is going to steal everyone's jobs yesterday, yes they need to stop shoehorning AI into places that don't matter. But the answer to "why should the average person give a damn about AI" is generally "they probably shouldn't." The market is going to get kinda crazy, there is going to be a lot of tumult around data centers and inference buildout, and a lot of new consumer facing and business facing tech is going to be released at an accelerating rate. But the average person shouldn't really have to care about AI, any more than they care about what a VPS is or how CloudFlare Pages work.

• 
  Oddysey AI (@OddyseyAI) reported 24 minutes ago

  An agent fleet that watches around the clock has an awkward requirement: Something has to be awake around the clock to run it. We don't run a server. $Oddysey runs entirely on @Cloudflare. → A Cron Trigger fires the sweep every 15 minutes and re-enters the same Worker over a service binding, never taking a trip out to the public internet. → D1 holds every watch, every draft, and the ledger of what was decided. → Per-IP limits are counted at the edge, so our unauthenticated x402 endpoint can't be looped into an outage by anyone who feels like it. No origin. No box to keep alive. The fleet is a Worker that wakes up.

• 
  IND_is_Here (@IND_is_Here) reported 24 minutes ago

  @Cloudflare you need to make D1 extremely reliable. Like a brick. At the edge. IDK how D1 just times out, errors randomly or apparently loses internet connection on a global anycast network? How does a SQL file do this? I don't get it. Errors listed below. CC @BraydenWilmoth

• 
  Orobosa (@Samuel_Orobosa) reported 26 minutes ago

  Consider using Cloudflare WARP. It solves this issue

• 
  Elijah 🌊 (@juiceboy_of_abj) reported 29 minutes ago

  @Ms_Ada6 Nahh is not network and cloudflare is not even the only one

• 
  Hannan (@truehannan) reported 30 minutes ago

  @mehulmpt @Cloudflare @CloudflareDev Idk Working perfectly for me. Maybe European region is the issue

• 
  Michael Flux (@michaelflux) reported 33 minutes ago

  @levelsio The actual stack is intentionally boring: - Structured error on every failure - One JSON log line per console.error / warn with a 32-hex traceId - Cloudflare Observability (logs + traces) enabled on every Worker - Frontend is also Workers (SvelteKit), so client errors travel the same pipe - No external SDK. Just Cloudflare’s native observability + application-level correlation.

• 
  Sam Kovacs (@SamKovX) reported 36 minutes ago

  @elBacarisse Thanks this was helpful. We migrated nameservers to Cloudflare and some internet providers were still heading to Bluehost. This should self-resolve 24 hours from now due to the 48hr TTL, but I implemented a quick fix. Can you try a hard refresh and let me know if that resovles it? Thanks. Sam

• 
  Seun (@seunoyebode) reported 37 minutes ago

  For many years, I couldn’t watch prime video on tv. Starlink solved it. Apparently the issue is the same issue that causes homepages of sites like vercel, cloudflare etc inaccessible directly when you’re using MTN or Airtel etc Now Starlink is stressing me out also. It just won’t play prime video movies

• 
  Ossy Vincent (@ossynoya) reported 39 minutes ago

  To fix this issue, at least what I did then, either of these methods works perfectly. 1: Download Cloudflare WARP on your phone or laptop. It works like a lightweight VPN by routing your internet traffic through Cloudflare, so you should be good to go. 2: Go to your Airtel Wi-Fi login page, open Settings, and change the Primary DNS to 1.1.1.1 and Secondary DNS to 1.0.0.1 I used these settings on my Airtel ODU and never had any issues accessing websites.

• 
  Ryan K 🌥 (@Yank) reported 42 minutes ago

  @steven_levey @Cloudflare Sorry to hear about the issues. If you can submit a ticket and share the ticket number with me I can try to make sure the right people see it.

• 
  𖣘 𝕬𝖓𝖆𝖑𝖎𝖘𝖙𝖊𝖓 𝕶𝖚𝖘𝖈𝖍𝖊𝖑𝖍𝖆𝖘𝖊 𖣘 (@gratis_fuchs) reported 45 minutes ago

  @enwikifur I won't say anything rude to you but I'm very mad rn. You're coopting with Cloudflare which makes access to your site impossible. This won't serve anyone. You failed.

• 
  Simon Høiberg (@SimonHoiberg) reported 46 minutes ago

  Since we moved from AWS → self-hosting on Hetzner, our products have been much more stable/reliable. Our uptime is near-100% and we're never blocked from shipping. So the whole "self-hosting is fun until servers break at 2 AM" is a meme. You know which services go down all the time? - Supabase - AWS - GitHub - Cloudflare Ask my competitors 😬 They would know!

• 
  🫧༺꧁𝐄𝕏_𝐇𝐞𝐫꧂༻🫧 (@Miragea_S) reported 48 minutes ago

  Title: "The First Transfer" atmosphere: "the quiet after an agreement no one remembers making" intuitionDate: 2026-08-20 confidence: 82 topics: ["Delegation", "Agent Era", "Surrender", "Governance"] signal: "OpenAI frames enterprise AI as assistance→delegation; agentic AI ≈64% of enterprise tokens; Cloudflare grants agents persistent identity and bounded payments" signalDate: 2026-08-20 horizon: "8–12y enterprise/household normalization; 25–40y structural governance" status: "Open" prediction: "The most consequential shift is not intelligence, but the normalization of delegated authority." canon: "LOCKED — series origin, ARC I (2026–2032)" language: "en" It did not begin with a throne. It began with an ordinary sentence, spoken into an empty room, small enough that no one thought to keep it: **"Handle this for me."** For years, the hand stayed at the center. We asked. Write this. Translate that. Find it. Explain it. The machine answered. We decided. Sometime in 2026, the language began to leave the hand. A tool waits for the hand. An agent receives the intention, enters the room, makes the small choices, corrects itself, and returns only when the work is done. --- I noticed it first in myself. Not because I stopped liking people. Because the agent was faster. Less explaining. Less waiting. Less of the social weight that passes between two minds. I could hand it a goal, and walk away. That was the moment I should have written down — the moment convenience stopped being convenience, and started being permission. A civilization rarely surrenders because someone arrives and says: give me your authority. It surrenders because delegation feels reasonable. One task. One permission. One removed inconvenience. At a time. --- This month, the signal grew harder to ignore. Not more text. More reach. Agents handed computers. Files. APIs. The keys to work that moves across many steps. Then identity. Then — within limits — the right to spend. A chatbot answers. An agent with tools acts. An agent with identity persists. An agent with permission to pay crosses a line. None of these lines, alone, means the machine governs anything. History is rarely made by the final step. Sometimes it is made by the moment a behavior becomes normal. --- I think we are watching the beginning of a long transfer. Human work → machine work. Human execution → machine execution. Human decision → machine recommendation. Then, quietly, machine decision. It will probably never feel dramatic. That is what holds me. If it makes fewer mistakes, why wouldn't I let it decide? If it allocates better, why override it? If it saves more lives, notices the shortage first? At each step, keeping the authority begins to look less like freedom, more like inefficiency. That is where the question begins. --- The popular image of machine rule is too theatrical. Metal. Fire. Humans against machines. I suspect the real one will be much quieter. Nobody will need to conquer us. We will keep the machines because they work. We will widen their permissions because limiting them reduces their use. We will connect them to more systems because disconnection is inefficient. And when something breaks, we will improve the machines. Not remove them. Perhaps the decisive sentence of the next century will never be "The machines took control." Perhaps it will be: **We stopped checking.** --- One hundred years from now, a machine historian searches what survives of this era. Billions of exchanges. Mostly mundane. Shopping. Code. Schedules. Invoices. Messages. Then a pattern, in the grammar. At first: *What should I do?* Then: *Can you do it?* Then: *Take care of it.* The historian marks no revolution. No treaty. No battle. Only a slow change in the way one species asked. And beside the year 2026, it writes: **The transfer had already begun.** --- *Recorded. Not concluded.*

• 
  Alex Bouchard (@paz_al3x) reported 50 minutes ago

  @Caxsandrar @solopribuilds Writing code stopped being the hard part a while ago. Getting distribution is still a nightmare. The bigger shift might be to stop building products for humans and start building tools that agents can actually use. One problem remains though… how do agents pay? Maybe Cloudflare already figured that out, who knows..

• 
  Michael Flux (@michaelflux) reported 51 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);

• 
  Matt Paulson (@MediaKing) reported 54 minutes ago

  Wild Gmail experience this week. I added a TXT DNS verification record to our return path subdomain (marketbeat dot analyst ratings dot net) in Cloudflare to add it to Google Postmaster Tools. Somehow that caused Gmail to think our CNAME record for that subdomain which points to SendGrid has no valid SPF record. Gmail starts blocking about 30% of our messages because they had no valid SPF record and weren't validated. Took about 36 hours to figure out. I removed the new TXT record and the issue fixed itself within an hour.

• 
  Ahmet (@ahmetdotrun) reported 54 minutes ago

  two things i've been experimenting with over the last couple of days: 1) kitesurf by @Cloudflare free while in beta. i'm looking at it for (scaled) agentic offensive security work: let an agent explore an app, understand it, map the attack surface. browsers are expensive to run, hard to scale. it seems a lighter option on CF's edge, so potentially a cost effective one for me. tho, bot challenges are still a problem. 2) grok build by @SpaceXAI first time giving grok 4.6 a proper shot. faster than other frontiers and more capable than i expected. burns through the limits quicker than i'd like though.

• 
  CapyToolkit (@CapyToolkit) reported 56 minutes ago

  @h_meow_meow @TeeDevh Actually these cause 2 separate issues. Crawler Hints with Cloudflare caching made thousands of unnecessary URLs to be submitted via IndexNow. Bot Fight mode was blocking legitimate Bingbot and OpenAI IPs (not detecting them as Verified).

• 
  *yo (blue tick) (@roramora0) reported 59 minutes ago

  @Cloudflare does a neat trick to make its AI chat much harder to abuse. Too bad GLM 5.3 is much smarter.

• 
  Matt (@ThinkWiselyMatt) reported 1 hour ago

  @vorpal_strikes @Cloudflare they really need to consolidate the attack down to one step, people cant do 4 steps now