---
title: "Proxy server: meaning and definition"
url: https://proxy.wiki/glossary/proxy-server/
type: Glossary Term
author: "proxy.wiki editorial"
published: 2026-08-19
updated: 2026-10-21
site: proxy.wiki
topics: ["Proxy fundamentals"]
license: CC BY 4.0; quote freely with attribution to https://proxy.wiki/
---

# Proxy server: meaning and definition

> A server that makes requests on your behalf, so the destination sees the proxy's address instead of yours.

The HTTP standard defines it more tightly than everyday speech does. RFC 9110 calls a proxy a message-forwarding agent “chosen by the client, usually via local configuration rules”. The word doing the work is _chosen_. You point your software at a proxy on purpose, and that is what separates it from a [reverse proxy](/glossary/reverse-proxy/), which a site’s operator puts in front of their own servers, and from a [transparent proxy](/glossary/transparent-proxy/), which a network forces on you without asking.

This entry keeps to the definition and the evidence for it. The full introduction, with the four proxy types, authentication and how to pick a first proxy, is the guide [What is a proxy server?](/guides/what-is-a-proxy-server/).

## The bytes a client sends to a proxy

A proxied request looks different on the wire from a direct one, and the difference is visible in the first line. On 21 October 2026 we pointed curl 8.7.1 at a small logging proxy on 127.0.0.1 and printed the request head it received, with each line ending shown as `\r\n`. For a plain `http://` URL, curl sent the whole URL, not just the path:

```
GET http://example.com/ HTTP/1.1\r\n
Host: example.com\r\n
User-Agent: curl/8.7.1\r\n
Accept: */*\r\n
Proxy-Connection: Keep-Alive\r\n
\r\n
```

That full URL on the request line is the “absolute-form” that RFC 9112 requires a client to use when talking to a proxy. A server that receives it knows it is being asked to forward. For an `https://` URL, with credentials in the proxy URL, curl sent something shorter and stopped:

```
CONNECT example.com:443 HTTP/1.1\r\n
Host: example.com:443\r\n
Proxy-Authorization: Basic dXNlcjpwYXNz\r\n
User-Agent: curl/8.7.1\r\n
Proxy-Connection: Keep-Alive\r\n
\r\n
```

The proxy accepts with a 2xx status (ours sent `HTTP/1.1 200 Connection established`), and from then on it relays encrypted bytes it cannot read. It learned a hostname and a port. It never sees the path, the headers or the body of what follows.

Two details in those captures are worth knowing. The `Proxy-Authorization` value is Base64, not encryption: `echo dXNlcjpwYXNz | base64 -d` prints `user:pass`, so anyone who can read the connection between you and the proxy can read the credentials. And curl still sends `Proxy-Connection: Keep-Alive`, a header RFC 9112 says clients are “encouraged not to send” because it never worked reliably across layered proxies. It is harmless, but it is one of several small marks that identify curl to whatever sits on the other end.

To reproduce the capture without writing any code, listen with netcat and point curl at it from a second terminal. Netcat prints the head and curl waits for a reply that never comes. Stop netcat and curl gives up with exit code 56, which its manual lists as a failure receiving network data. The message it printed for us was `curl: (56) Proxy CONNECT aborted`.

```
nc -l 8899
curl -x http://user:pass@127.0.0.1:8899 https://example.com/
```

## What each kind of proxy request reveals

The request form decides how much the proxy operator can see. The same split applies to SOCKS5, which works below HTTP and carries a host and port rather than a URL.

| Request form | Used for | The proxy learns | The proxy cannot see |
| --- | --- | --- | --- |
| Absolute-form, such as GET http://host/path | Plain HTTP URLs | The full URL, every header, the body | Nothing in the request |
| CONNECT host:443 | HTTPS through an HTTP proxy | Hostname, port, timing, byte counts | Path, query, headers, body |
| SOCKS5 CONNECT with a domain name | [socks5h://](/glossary/socks5/) in curl | Hostname, port, timing, byte counts | Everything inside the TLS session |
| SOCKS5 CONNECT with an IP address | socks5:// in curl | Destination IP and port, plus the hostname from the TLS server name (SNI) your client still sends | Everything inside the TLS session, and the DNS lookup your machine made itself |

The last row is the reason `socks5://` leaks DNS lookups to your own network while `socks5h://` does not. Resolving locally does not hide the hostname from the proxy, though: the TLS ClientHello names the server in plain text unless both ends use Encrypted Client Hello. The [curl guide](/guides/how-to-use-a-proxy-with-curl/) covers that trap and the flags around it.

A proxy can hide your address from the destination. It never hides it from the proxy. In every row above the operator knows your address and which host you reached. Choosing a proxy is a decision about whom you trust with that.

## What it is not

A proxy is configured per application. A VPN moves the whole device’s traffic at the operating-system level, so an application that ignores the proxy setting still goes out through the VPN, but goes out directly when only a proxy is set. Home routers doing NAT also substitute an address, yet nobody chose them as an intermediary and they do not read HTTP, so they are not proxies in the standard’s sense.

Never assume a proxy setting took effect. Ask an address-echo service what it saw, then repeat without the proxy as a control, as the [curl guide](/guides/how-to-use-a-proxy-with-curl/) shows. Clients also differ in which settings they read: [HTTP_PROXY, HTTPS_PROXY and NO_PROXY](/guides/proxy-environment-variables/) behave differently in curl, Python and Node.

## Frequently asked questions

### Can the proxy operator read my requests?

For plain HTTP, yes, all of it: the full URL, every header and the body. For HTTPS the operator sees the hostname, the port, the timing and the byte counts, because the request travels inside a CONNECT tunnel that the proxy relays without decrypting.

### Is a proxy the same as a VPN?

No. A VPN moves every application's traffic at the operating-system level. A proxy is set per application, so a program that ignores the setting connects directly while the others go through the proxy.

### Does a proxy hide my traffic from my internet provider?

Partly. Your provider sees that you connect to the proxy and how much data you move. If the proxy URL is plain http://, it can also read the CONNECT line naming each site and the Base64 credentials. An https:// proxy URL encrypts that leg as well.

## Sources

1. [RFC 9110 section 3.7: Intermediaries (proxy, gateway, interception proxy)](https://www.rfc-editor.org/rfc/rfc9110.html#section-3.7)
2. [RFC 9110 section 9.3.6: CONNECT](https://www.rfc-editor.org/rfc/rfc9110.html#section-9.3.6)
3. [RFC 9112 section 3.2.2: absolute-form request target](https://www.rfc-editor.org/rfc/rfc9112.html#section-3.2.2)
4. [RFC 9112 appendix C.2.2: Keep-Alive connections and Proxy-Connection](https://www.rfc-editor.org/rfc/rfc9112.html#appendix-C.2.2)
5. [curl manual: proxy options and exit codes](https://curl.se/docs/manpage.html)
6. [RFC 6066 section 3: Server Name Indication](https://www.rfc-editor.org/rfc/rfc6066.html#section-3)
