developers

curl exit 5 means the proxy's hostname did not resolve, usually a leftover variable or a typo. Exit 97 is a failed proxy handshake, usually SOCKS. The fixes for both.

Docker has three proxy settings: the daemon's, the client's for containers, and Docker Desktop's. Which one each error points at, and the certificate fix.

How git finds its proxy, how to set and remove it, and the fix for each error: could not resolve proxy, 407 after CONNECT, and the SSL certificate problem.

Headless Chromium reports proxy failures as net:: errors on page.goto. What each means, why 127.0.0.1 fails inside Docker, and how each framework passes credentials.

npm's network errors all end with the same proxy hint. How npm finds its proxy, the .npmrc keys, the certificate fix for corporate proxies, and what each error means.

How pip finds its proxy, the --proxy flag and pip.conf keys, and the fix for each error: cannot connect to proxy, 407 in the tunnel, and certificate verify failed.

Every AI coding tool, Claude Code, Cursor, Copilot, Cline and the rest, raises the same proxy question, and the answer is always two answers. The tool's own connection to its model rarely needs a proxy, and never to hide you. The code and agents you build with it often do. This explainer draws the line once, so the per-tool guides can skip it.

Proxies for Claude Code split into three questions. The big one is the code it writes for you, scrapers, crawlers, monitors and browser agents, which needs residential proxies to run at scale. The second is routing Claude Code itself through a proxy, which it documents with HTTP(S)_PROXY (no SOCKS) for corporate networks and reachability. The third is everything a proxy never does.

Cline is unusual among AI coding agents: it has a built-in browser tool, so it browses the web itself, and it builds automation that does the same. Both meet the walls a proxy solves. This covers where the proxy goes for the code Cline writes, the reachability angle of its bring-your-own-key model access, and the honest limits.

Cursor is an agent-first IDE, and its agents write and run the scrapers, crawlers and automation that need proxies. This covers the residential-proxy layer for what Cursor builds, Cursor's own VS Code proxy settings for corporate networks, and what an IP never does for the editor itself.

GitHub Copilot's proxy story is mostly the corporate one: it documents proxy settings per IDE, basic and Kerberos authentication, and custom certificates for TLS-inspecting networks. That is your company's proxy, not a residential one you buy. The residential proxy belongs to the scrapers and agents you build with Copilot, and this covers both.

An MCP server that fetches the web, a scraper, a search tool, a price checker, hits the same bot walls and per-IP limits as any client, so it needs residential proxies. One that talks only to a private API does not. This covers which MCP servers need a proxy, where it goes in the server, and how to wire it in the common runtimes.

The Claude Agent SDK is Claude Code as a library, so the agents you build with it browse the web through built-in and custom tools. This guide shows where a proxy goes: in the custom tools and subprocess code that fetch sites, and the process-level HTTP(S)_PROXY for reachability, with the honest note that the SDK's own model calls do not want one.

Free proxies for Scrapy are fine for testing your middleware, not a real crawl. The rotating-middleware config, why dead exits thrash a run, and how to verify.

Free proxies for Selenium work for a throwaway session, not an unattended run. How to set the Chrome proxy arg and Selenium Wire, and why free exits hang.

Roblox blocks HttpService from calling roblox.com. Here is why, how RoProxy and self-hosted proxies solve it, plus the measured per-endpoint rate limits and batch ceilings that decide your real throughput.

Do free proxies for web scraping actually work? Where the limits are, the safest free options, how to test them, and when to upgrade to reliable proxies.

Proxies for Scrapy explained: which proxy type fits your target, real settings.py and middleware setup, sticky vs rotating, and how to stop bans and 429s.

Three cURL exit codes cover almost every proxy failure: 7, 56 and 60. What each one means, why proxies trigger it, and the real fix, with copy-paste commands.

Every way to route curl through a proxy: -x syntax, SOCKS5 vs socks5h, proxy auth, env vars, timing flags, plus scripts to test whole proxy lists in parallel.

How to choose and use proxies for web scraping: proxy types ranked by job, rotation strategy, avoiding blocks, request hygiene, and code you can run today.