What it means: The client started TLS, but the server answered with something that is not TLS, usually plain HTTP.
https_proxy uses http://ssl to the listen lineFree online tool — SSL certificate checker: instant results, no signup.
Runs a full TLS check: handshake, protocol versions, certificate chain, trust and expiry.
OpenSSL reason 267, SSL_R_WRONG_VERSION_NUMBER, string wrong version number. Every TLS record starts with a type byte and a protocol version; when the first bytes the client receives do not look like a TLS record header, OpenSSL stops with this reason. The answer it got was not a TLS ServerHello at all. The most common thing on the other end is a plain HTTP server, whose HTTP/1.1 400 or HTTP/1.1 200 starts with bytes that OpenSSL tries to read as a version.
The packed error code tells you the OpenSSL generation. OpenSSL 3 prints error:0A00010B:SSL routines::wrong version number (library 20 shifted into the top bits, reason 0x10B); OpenSSL 1.1.1 printed error:1408F10B:SSL routines:ssl3_get_record:wrong version number (sslerr.h, ssl_err.c). You meet it as:
curl: (35) ... wrong version numberssl.SSLError: [SSL: WRONG_VERSION_NUMBER] wrong version numbercode: 'ERR_SSL_WRONG_VERSION_NUMBER'Despite the name, it is almost never about TLS 1.2 versus 1.3. A real version disagreement produces tlsv1 alert protocol version or a handshake failure instead.
| Cause | Example | Fix |
|---|---|---|
| https:// to a plain-HTTP port | https://example.com:8080/ where the app listens without TLS | Use http:// for that port, or terminate TLS in front of it |
| Proxy URL with the wrong scheme | https_proxy=https://proxy:3128 for an ordinary HTTP proxy | Write https_proxy=http://proxy:3128 |
| Server listens on 443 without TLS | nginx listen 443; with no ssl parameter | listen 443 ssl; |
| Wrong mail port mode | Implicit TLS on 587, or STARTTLS on 465 | 465 = TLS from the first byte, 587 = STARTTLS |
| Something in the path answers in clear text | Captive portal, transparent proxy, load balancer health page | Test from another network; fix the device |
The proxy case catches many people because the variable is called https_proxy. It names the proxy used for HTTPS requests, and the scheme in its value describes how to talk to the proxy. curl treats an https:// proxy as an HTTPS proxy and starts TLS to it (curl manual, --proxy); a Squid listening in plain HTTP on 3128 answers with HTTP, and you get wrong version number before the request even leaves for the real site.
The mail case follows RFC 8314: port 465 carries mail submission over implicit TLS, while 587 starts in clear text and upgrades with STARTTLS (RFC 8314). Python's smtplib.SMTP_SSL on 587 fails with exactly this error.
curl -v http://example.com:443/
# < HTTP/1.1 200 OK -> the "TLS" port is answering in plain HTTP
openssl s_client -connect example.com:443 -servername example.com </dev/null
# ... SSL routines::wrong version numberIf plain HTTP to the TLS port works, the server is misconfigured. If the server is fine from your laptop but not from a CI runner or office machine, look at the environment of the failing one:
env | grep -i _proxy
curl -v --noproxy '*' https://example.com/Success with --noproxy '*' means the proxy setting is the cause. For mail servers, test both modes explicitly:
openssl s_client -connect mail.example.com:465 -brief </dev/null
openssl s_client -connect mail.example.com:587 -starttls smtp -brief </dev/nullhttps:// only on ports where TLS is configured. An app server on 8080 or 3000 usually expects http:// unless you set up TLS on it.http_proxy and https_proxy to http://proxy.example.com:3128 unless your proxy really accepts TLS on that port. Check /etc/environment, CI variables, Docker daemon proxy settings and ~/.npmrc as well as the shell.ssl parameter: listen 443 ssl; (and listen [::]:443 ssl;). Run nginx -T | grep -n 'listen' to find a 443 listener without it, often a second server block that wins for some names.https:// gets this error.SMTP_SSL, secure: true) and 587 with STARTTLS (starttls(), secure: false).openssl s_client -connect example.com:443 -servername example.com -brief </dev/null
# Protocol version: TLSv1.3
curl -sS -o /dev/null -w '%{http_code}\n' https://example.com/
# 200The SSL/TLS checker connects to port 443 of your host and reports the negotiated protocol and certificate, so a plain-HTTP listener on 443 shows up immediately.
SSL/TLS is the encryption protocol that protects data between the browser and server. Our tool analyzes the certificate, chain of trust, TLS version, and knownvulnerabilities.
Issuer, validity period, signature algorithm, covered domains (SAN), and validation type (DV/OV/EV).
Full chain verification: from leaf certificate through intermediates to root CA.
Protocol version (TLS 1.2/1.3), cipher suites, Perfect Forward Secrecy (PFS) support.
Set up a monitor — get Telegram and email alerts 30/14/7 days before expiration.
SSL certificate monitoring
TLS config audit
HTTPS as ranking factor
customer trust
www and subdomains.Strict-Transport-Security header forces browsers to always use HTTPS.SSL certificate monitoring, check history and alerts 30 days before expiry.
Sign up freeAlmost never. It means the reply was not TLS at all, typically plain HTTP on the port you treated as HTTPS. A real version mismatch gives a protocol version alert instead.
Your https_proxy variable probably starts with https:// although the proxy speaks plain HTTP. Change it to http://proxy-host:port.
It is OpenSSL 3's packed code for the SSL library and reason 267, wrong version number. In OpenSSL 1.1.1 the same reason appeared as 1408F10B.
Port 465 uses implicit TLS from the first byte; port 587 starts in plain text and upgrades with STARTTLS. Mixing them up gives wrong version number.
Yes. An nginx listen 443 line without the ssl parameter serves plain HTTP on the HTTPS port, and every TLS client fails with wrong version number.
Free plan — 10 monitors, checks every 5 min, no card required. Upgrade for 1-minute interval and multi-region monitoring.