Blog
Troubleshooting· 4 min read

The program will not take a proxy login and password: which clients cannot send them, and what to do

There is a class of failures where the password is right and the proxy is alive, yet nothing connects: the client physically cannot send credentials. From the proxy's side that looks exactly like a wrong password, which is why people spend weeks changing the password when what needs changing is the authentication method.

In this article
  1. Chrome and Chromium: yes over HTTP, never over SOCKS5
  2. Who else cannot send a password
  3. Why it works in Firefox and not in Chromium
  4. The check: proxy or client
  5. The fix: authenticate by IP
  6. When a whitelist is the wrong choice
  7. About the local relay

Chrome and Chromium: yes over HTTP, never over SOCKS5

This is written down in Chromium's own documentation. On HTTP proxies: “HTTP proxies in Chrome support the same HTTP authentiation schemes as for target servers: Basic, Digest, Negotiate, NTLM” — the typo is in the original. On SOCKS5, the same document says outright: “No authentication methods are supported for SOCKSv5 in Chrome (although some do exist for the protocol).”

So it is neither your connection string nor the proxy: the support is absent from the browser. Asked on the chromium-dev list how to connect to SOCKS5 with authentication, a networking maintainer answered in one line: “This is not supported.”

Who else cannot send a password

  • Electron applications: the docs on --proxy-server state “The proxy URL does not support username and password authentication”.
  • Android's system proxy: the ProxyInfo class exposes only host, port and an exclusion list — there is no credential field in the API at all, and no protocol choice either.
  • netsh winhttp on Windows: the set proxy syntax takes only a server address and a bypass list; no credential parameter exists.
  • The manual proxy screen in Windows Settings: there is no password field on it.

And the ones that handle it fine: Firefox (it prompts for the login and password itself), curl (--proxy-user), iOS (the Wi-Fi settings expose an Authentication toggle with username and password), and macOS, where credentials are first-class arguments of the networksetup command.

bash
# macOS: login and password are ordinary arguments
networksetup -setsocksfirewallproxy Wi-Fi GATEWAY PORT on LOGIN PASSWORD

# Windows: there is no credential parameter here at all
netsh winhttp set proxy proxy-server="GATEWAY:PORT"

Why it works in Firefox and not in Chromium

Firefox implements SOCKS5 username and password authentication; Chromium does not. That is the source of the most common puzzle in antidetect browsers: the same proxy string works in one tool and not in another. The cause is not the tool but its engine — Chromium-based browsers have to build their own layer over the proxy for authenticated SOCKS5 to work at all, while Firefox-based ones get it from the engine.

The check: proxy or client

Before changing settings, separate the two. curl can pass credentials independently of the address, so its answer speaks only about the proxy.

bash
# socks5h, not socks5: the h means the proxy resolves names —
# which is what a browser does, so the DNS path matches production.
curl -x socks5h://GATEWAY:PORT -U LOGIN:PASSWORD https://example.com -o /dev/null -w '%{http_code}\n'

# The same credentials on the HTTP port:
curl -x http://GATEWAY:PORT -U LOGIN:PASSWORD https://example.com -o /dev/null -w '%{http_code}\n'

# 200 on both -> proxy and credentials are fine, the client is the problem.
# 407 -> the credentials are the problem, the client is irrelevant.

The fix: authenticate by IP

If a client cannot send a password, stop requiring one. With address-based authentication the machine's IP goes on a whitelist and the proxy recognises the client by the connection itself — only host and port remain in the settings, and the password field stays empty. This is not a workaround but the second supported authentication method.

Both methods are active at once here, so nothing has to be switched and no support ticket is needed: a server runs on the whitelist while scripts on a laptop keep using login and password.

When a whitelist is the wrong choice

  • A home address from an ISP changes — the whitelist then needs updating, and the job will stop at the least convenient moment. For those cases the address is updated with an API request.
  • A shared or NAT'd network: other people sit behind the same external address, and a whitelist opens access to all of them. Here you want a login and password, and a client that can send them.

About the local relay

The workaround forums recommend: run a local proxy without authentication that upstreams to ours with credentials. It works, but it is an extra hop and a second point of failure — when something breaks you now debug two places instead of one. With a whitelist available there is usually no reason for it.

Both authentication methods at once

Login with password and IP whitelist run in parallel on every plan — there is nothing to choose between. Rotating IPv6 from 650 ₽ for 50 threads, static IPv4 from 300 ₽.

View pricing

Proxies for this job

Check it with our tools