Localhost not working?

ERR_CONNECTION_REFUSED · ECONNREFUSED · Firefox: “Unable to connect”

The request reached your computer, but no program is listening on that port. Work through these in order:

1. Is the server still running? Go back to the terminal where you started it. If it crashed, the error is near the bottom; if you closed the window or your laptop slept, it's simply gone. Start it again — npm run dev, python manage.py runserver, docker start my-app — and wait until it prints an address.
2. Are you on the right port? Use the exact URL the server printed. Many tools quietly move to the next free port (“Port 3000 is in use, trying 3001 instead”).
3. Is anything listening at all? Ask your OS directly. No output means nothing is behind that door.
4. Try 127.0.0.1 instead of localhost. If that works, jump to the IPv6 mismatch.

# Is anything listening on port 3000?
lsof -nP -iTCP:3000 -sTCP:LISTEN

# Talk to it directly — "Connection refused" confirms it
curl -v http://localhost:3000

# Is anything listening on port 3000?
ss -ltn 'sport = :3000'

# Talk to it directly — "Connection refused" confirms it
curl -v http://localhost:3000

# Is anything listening on port 3000?
Get-NetTCPConnection -LocalPort 3000 -State Listen

# Talk to it directly
curl.exe -v http://localhost:3000

Error: listen EADDRINUSE: address already in use :::3000 · OSError: [Errno 48] Address already in use

Only one program can listen on a port. Another one — very often an old copy of your own server running in a forgotten terminal — already has it. Find it and stop it, or start your app on a different port.

# Find the process holding port 3000
lsof -nP -iTCP:3000 -sTCP:LISTEN

# Stop it (PID is the second column)
kill <PID>
# Still there? Force it
kill -9 <PID>

# Find the process holding port 3000
sudo ss -ltnp 'sport = :3000'

# Stop it — or kill whatever holds the port in one step
kill <PID>
sudo fuser -k 3000/tcp

# Find the process holding port 3000 (OwningProcess is its PID)
Get-NetTCPConnection -LocalPort 3000 -State Listen

# Stop it
Stop-Process -Id <PID> -Force

Or leave it running and pick another port: npm run dev -- --port 5174 (Vite), next dev -p 3001, flask run --port 5001, python manage.py runserver 8001.

connect ECONNREFUSED ::1:3000 · works in one tool, refused in another

localhost can resolve to two addresses: 127.0.0.1 (IPv4) and ::1 (IPv6). If your server listens on only one of them and your browser or tool tries the other, the connection is refused — even though the server is running. Node.js 17 and later stopped preferring IPv4 for localhost, which made this a common surprise.

curl -4 http://localhost:3000   # IPv4: 127.0.0.1
curl -6 http://localhost:3000   # IPv6: ::1

Fix: point your client at the address that works (http://127.0.0.1:3000), or make the server listen on both — for example server.listen(3000, "::") in Node.js accepts IPv6 and, on most systems, IPv4 too.

Phone shows “can't connect” · another computer on the same Wi-Fi can't reach it

On your phone, localhost means the phone. To reach your laptop you need three things:

1. Your laptop's LAN address, like 192.168.1.23 — not localhost.
2. A server listening on 0.0.0.0, not only 127.0.0.1 (Vite: npm run dev -- --host; Flask: --host=0.0.0.0; Django: runserver 0.0.0.0:8000).
3. The same network, and a firewall that allows it. Guest and office Wi-Fi often isolate devices from each other.

ipconfig getifaddr en0

hostname -I

ipconfig

Then open http://192.168.1.23:3000 (with your address and port) on the phone. See who can reach what, interactively →

docker ps shows “Up” · browser shows refused or an empty response

Two things must both be true. The port must be published from the container to your computer with -p HOST:CONTAINER, and the app inside must listen on 0.0.0.0 — because inside a container, 127.0.0.1 means the container itself.

# Wrong: the container runs, but no port is published
docker run my-app

# Right: your localhost:8080 → the container's port 80
docker run -p 8080:80 my-app

# Check the PORTS column for 0.0.0.0:8080->80/tcp
docker ps

With Docker Compose, the same mapping goes under ports: ["8080:80"]. Going the other way — reaching a server on your computer from inside a container — use host.docker.internal instead of localhost (Docker Desktop).

curl works inside WSL · Edge or Chrome on Windows can't connect

WSL 2 runs Linux in a lightweight virtual machine with its own network. Windows normally forwards localhost into it automatically; when that fails, make the server listen on 0.0.0.0, try the WSL address from hostname -I, or switch WSL to mirrored networking on Windows 11, which shares your Windows network with Linux:

[wsl2]
networkingMode=mirrored

Then run wsl --shutdown and start your distribution again.

https://localhost:3000 fails · NET::ERR_CERT_AUTHORITY_INVALID · ERR_SSL_PROTOCOL_ERROR

Most dev servers speak plain http://. If your browser autocompletes or upgrades to https://, there's nothing on the other end that speaks encryption. Type http://localhost:3000 explicitly.

If Chrome keeps forcing HTTPS for localhost (an old HSTS rule), open chrome://net-internals/#hsts and delete the domain security policy for localhost. And if you genuinely need HTTPS locally, create a certificate your machine trusts:

mkcert -install
mkcert localhost 127.0.0.1 ::1

ERR_CONNECTION_TIMED_OUT · the tab spins forever

A refusal is instant; a timeout means packets are being dropped or the server is stuck. Usual suspects: a firewall, VPN, proxy or antivirus intercepting local traffic; a server frozen in a busy loop; or an IP address that isn't yours anymore because you changed Wi-Fi.

curl -v --max-time 5 http://localhost:3000

Pause the VPN or security tool and try again. If that fixes it, allow your port in its settings. If the server itself hangs, restart it and watch its logs.

White screen · changes don't show up · works in a private window

A blank page is often a crashed frontend: open the browser console (F12) and look for red errors. Old versions usually come from caches and service workers, which can keep serving a previous build even after your server changed.

• Hard refresh: ⌘ Shift R on macOS, Ctrl Shift R on Windows and Linux.
• DevTools → Application → Service workers → Unregister, then reload.
• Rule it out completely with a private window or another browser.

localhost resolves to the wrong address · works on every machine but yours

Your hosts file maps names to addresses before any DNS server is asked, and a tool or tutorial may have edited it. It should contain 127.0.0.1 localhost and usually ::1 localhost.

cat /etc/hosts
# Windows:
type C:\Windows\System32\drivers\etc\hosts

Fix any other line that mentions localhost (you'll need admin rights to save the file).

http://localhost:3000 works for you · “refused to connect” for everyone else

That's not a bug — it's what localhost means. On your friend's device, localhost is their computer, where nothing is running. There are two real fixes:

• Quick demo: a tunnel such as cloudflared tunnel --url http://localhost:3000 or ngrok http 3000 gives your laptop a temporary public URL. It stops working when your laptop sleeps — and anyone with the link can reach your app.
• The durable fix: deploy it to a server in the cloud that stays on and has a public address. Here's how, in one sentence.