FREE · Proxy for AI agents

A free proxy
for your AI agent.

Agents make lots of fast requests from one IP, and sites rate-limit per IP. Point your agent at one address and every page it reads leaves from a different IP. No signup to try: 1 GB a day per IP. With a free account: 2.5 GB a day, 25 GB a month.

870
Proxies live now
1 min
Re-test interval
25 GB
Free every month
$0
No card needed
Why agents get blocked

It's usually the IP, not your code

An agent on a laptop, Colab or a cheap VM sends every request from one address, with no human pauses. That address gets rate-limited first and banned second.

01

One address, no list

ws.solproxy.net:8501 picks a live proxy for each connection and retries on another if it dies. You never keep a list or weed out dead IPs.

02

Tested every minute

Every proxy behind it is re-tested every minute with a real HTTPS request. Proxies that tamper with TLS are dropped.

03

Sticky when the agent needs it

For a multi-step browser task, free-sticky10m-agent1 keeps the same exit for up to 10 minutes. For a country, free-de.

Drop it into your stack

Copy, paste, run

Any client that takes an HTTP proxy works. These are the ones agent builders ask about most.

web_mcp.py
# pip install "mcp>=2" httpx beautifulsoup4 import os, httpx from bs4 import BeautifulSoup from mcp.server.mcpserver import MCPServer TOKEN = os.getenv("SOLPROXY_TOKEN") # optional: works without it PROXY = f"http://free:{TOKEN}@ws.solproxy.net:8501" if TOKEN else "http://ws.solproxy.net:8501" mcp = MCPServer("web") @mcp.tool() async def fetch_page(url: str, max_chars: int = 12000) -> str: """Fetch a web page and return its readable text.""" for _ in range(4): # free proxies die: retry on a new IP try: async with httpx.AsyncClient(proxy=PROXY, timeout=30, follow_redirects=True) as c: r = await c.get(url) if r.status_code < 400: soup = BeautifulSoup(r.text, "html.parser") for tag in soup(["script", "style", "nav", "footer"]): tag.decompose() return soup.get_text("\n", strip=True)[:max_chars] except httpx.HTTPError: pass return f"Could not fetch {url}" if __name__ == "__main__": mcp.run() # stdio # Claude Code: claude mcp add web -e SOLPROXY_TOKEN=... -- python web_mcp.py

Your coding agent gets a fetch tool whose requests leave from the pool, not from your IP.

agent.py
# pip install browser-use import asyncio, os from browser_use import Agent, Browser, ChatOpenAI from browser_use.browser.profile import ProxySettings browser = Browser( headless=True, proxy=ProxySettings( server="http://ws.solproxy.net:8501", username="free-sticky10m-agent1", # same exit IP for up to 10 min password=os.environ["SOLPROXY_TOKEN"], ), ) agent = Agent( task="Open news.ycombinator.com and list the top 5 story titles.", llm=ChatOpenAI(model="gpt-4.1-mini"), # any LLM browser-use supports browser=browser, ) asyncio.run(agent.run())

A browsing agent does several steps on one site. Keep it on one IP with a sticky username; give each parallel agent its own session name.

crawl_tool.py
# pip install crawl4ai && crawl4ai-setup import asyncio, os from crawl4ai import AsyncWebCrawler, BrowserConfig, CacheMode, CrawlerRunConfig, ProxyConfig BROWSER = BrowserConfig(headless=True, proxy_config=ProxyConfig( server="http://ws.solproxy.net:8501", username="free", # flags go here: free-us, free-sticky10m-job1 password=os.environ["SOLPROXY_TOKEN"], )) RUN = CrawlerRunConfig(cache_mode=CacheMode.BYPASS, page_timeout=45_000) async def fetch_markdown(url: str, tries: int = 3) -> str: """Agent tool: fetch a page through the pool, return LLM-ready markdown.""" for _ in range(tries): # a new browser is a new IP async with AsyncWebCrawler(config=BROWSER) as crawler: r = await crawler.arun(url, config=RUN) if r.success and (r.status_code or 200) < 400: return r.markdown.raw_markdown raise RuntimeError(f"{url}: failed after {tries} tries") print(asyncio.run(fetch_markdown("https://news.ycombinator.com"))[:500])

Hand fetch_markdown to LangGraph, CrewAI, smolagents or any tool-calling API as "read this URL".

playwright_agent.py
import asyncio, os from playwright.async_api import async_playwright async def main(): async with async_playwright() as p: browser = await p.chromium.launch(proxy={ "server": "http://ws.solproxy.net:8501", "username": "free", "password": os.environ["SOLPROXY_TOKEN"], }) page = await browser.new_page() # keep certificate checks on await page.goto("https://api.ipify.org") print(await page.inner_text("body")) # the proxy's IP, not yours await browser.close() asyncio.run(main())

Skip the ignore_https_errors=True you see in proxy tutorials. HTTPS is a tunnel the proxy can't read into, as long as certificate checks stay on.

tool.py
import os, requests, httpx PROXY = f"http://free:{os.environ['SOLPROXY_TOKEN']}@ws.solproxy.net:8501" # requests r = requests.get("https://api.ipify.org", proxies={"http": PROXY, "https": PROXY}, timeout=30) print(r.text) # httpx (async agents) async def fetch(url: str) -> str: async with httpx.AsyncClient(proxy=PROXY, timeout=30) as c: return (await c.get(url)).text

No token? Use http://ws.solproxy.net:8501 as the proxy and drop the username and password.

Flags and limits

Exactly what you get for free

Flags go in the username, any order

UsernameWhat it does
freeNew proxy per connection
free-us, free-dePrefer that country; any country if none is live
free-sticky10m-job1Same proxy up to 10 min while it stays up (sticky1m to sticky1h)
free-residentialOnly IPs not owned by a known hosting company; 503 if none is live

Limits (reset 00:00 UTC)

No accountFree account
Per day1 GB / IP2.5 GB
Per month—25 GB
ProxiesPassed the latest checkPassed 9 of the last 10 checks
Connections16 at a time

The whole free pool also has a monthly budget; if it runs out, the gateway returns 503 until the 1st of the month. Paid traffic is not affected.

Before you build on it

What a free proxy fixes, and what it doesn't

These are public open proxies we find and re-test, not our residential network. They are good for a lot of agent work and useless for some.

✓ Helps with

  • Per-IP rate limits and "too many requests" errors
  • IP bans after your agent hammered one site
  • Datacenter blocks on Colab, GitHub Actions and VPS IPs
  • Docs, blogs, forums, listings and news pages

✕ Does not help with

  • Serious bot protection (Cloudflare challenges, DataDome, Amazon). In our own test the free pool got through 0 of 36 tries on protected sites.
  • Captchas
  • Logins. Never send a password or session cookie through a free proxy.
  • Uptime guarantees: free proxies die mid-request, so retry
When free is not enough

Same code, residential IPs that stay up

When you hit the daily cap or a target needs a real home IP: the paid pool is on the same host, port 8500. Same -de / -sticky10m flags, no subscription, no minimum.

$0.80/GB

pay as you go, down to $0.75/GB with packages

FAQ

Questions

How do I give my AI agent a free proxy?

Set the agent's HTTP proxy to http://ws.solproxy.net:8501. No account needed; that gives 1 GB a day per IP. With a free account, put your token where the password goes (http://free:TOKEN@ws.solproxy.net:8501) for 2.5 GB a day and 25 GB a month.

Do I need to sign up or add a card?

Not to try it. Without an account you get 1 GB a day per IP. The free account asks for no card.

Are these residential IPs?

The free pool is public open proxies we re-test every minute: some are home connections, some are datacenter. The free-residential flag only hands out IPs not owned by a hosting company. Our own residential network is the paid product.

Will it get my agent past Cloudflare?

Usually not. Serious bot protection looks at the browser fingerprint more than the IP; in our own test the free pool got through 0 of 36 tries on protected sites.

What happens when I hit the free limit?

The gateway answers 429 or 407 with the reason, and the limit resets at 00:00 UTC. If you can't wait, the paid pool is $0.80/GB, down to $0.75 with packages.

Want the raw proxies instead? Table, JSON and plain text: free proxy list.