AI Agent Proxy: Exit IPs for Browser Use, Playwright MCP and n8n

Why AI agents need their own proxy

AI agents now browse, call APIs and fill forms on their own, and sites have noticed: Cloudflare reported in July 2026 that more than half of Internet traffic is non-human, and since July 2025 it has blocked AI training crawlers by default on new domains. An agent that shares your production server's IP can get that IP rate-limited or blocked.

A proxy gives each agent a separate, controllable exit: a fixed datacenter IP for tool calls and APIs that allowlist callers, or a household IP in a chosen country when a task has to see the page a local visitor sees. It doesn't change who the agent is — say what it is and follow robots.txt and each site's terms.

Agent frameworks take a proxy directly. Browser Use's ProxySettings has server, bypass, username and password; Playwright MCP reads a proxy from its configuration file (its --proxy-server flag has no password option); n8n's HTTP Request node has a Proxy option that overrides HTTP_PROXY and HTTPS_PROXY. Browser-based agents run on Chromium, which can't authenticate to SOCKS5 — use the HTTP endpoint.

• Cloudflare blog — Content Independence Day, one year on (July 2026)
• Browser Use source — ProxySettings (server, bypass, username, password)
• Playwright MCP — README: --proxy-server and configuration file
• n8n docs — HTTP Request node: Proxy option
• Chromium — Proxy support in Chrome

Recommended AI agent setup

Scroll horizontally with the arrow keys to read all content.

The button below opens Static Datacenter Proxies with the AI and automation use case. Choose Region, Term, Protocol and Quantity (up to 100 IPs at a time) before you pay.

• Proxy302 console — Static Datacenter Proxies
• Chromium — Proxy support in Chrome

Which Proxy302 proxy to buy

Scroll horizontally with the arrow keys to read all content.

Prices in US dollars. Top up your wallet once and pay any line from it. Nothing renews automatically — renew in My assets to keep an IP.

• Proxy302 pricing

Set up proxies for AI agents in 5 steps

1. Choose the region and term

   The button below opens Static Datacenter Proxies (Per IP and term) with the AI and automation use case. Pick the Region, choose a Term and set Protocol to HTTP. Set Quantity to the number of IPs you need (one for each agent or concurrent session, up to 100 at a time). Click Review cost and generate.

2. Let your coding assistant wire it up

   In My assets, click Copy for AI on a proxy, pick the target — Python, Node.js, curl, Playwright, Puppeteer, Scrapy, AdsPower, BitBrowser or other — and paste the text into Claude Code, Cursor or ChatGPT. The text asks the assistant to configure the proxy in your project and verify the exit IP.

3. Or set it yourself

   Browser Use: pass ProxySettings to Browser (code below). Playwright MCP: put the proxy in the config file and start npx @playwright/mcp@latest --config playwright-mcp.json. n8n: open the HTTP Request node, add the Proxy option and enter the proxy URL.

4. Check the exit IP

   Ask the agent to open api.ipify.org and report the IP.

   Expected result The agent reports the proxy's IP, not your server's.

5. Keep secrets out of prompts

   Store the proxy password in an environment variable or a .env file excluded from git; don't paste it into the agent's task text.

Browser Use agent through an authenticated proxy (Python)

Scroll horizontally with the arrow keys to read all content.

# pip install browser-use   (model and API key setup: docs.browser-use.com)
import asyncio
import os
from browser_use import Agent, Browser, ChatOpenAI
from browser_use.browser.profile import ProxySettings

browser = Browser(
    proxy=ProxySettings(
        server=f"http://{os.environ['PROXY_HOST']}:{os.environ['PROXY_PORT']}",  # HTTP endpoint
        username=os.environ["PROXY_USER"],
        password=os.environ["PROXY_PASS"],
        bypass="localhost,127.0.0.1",
    )
)


async def main():
    agent = Agent(
        task="Open https://api.ipify.org?format=json and report the IP address it shows",
        llm=ChatOpenAI(model=os.environ["LLM_MODEL"]),
        browser=browser,
    )
    history = await agent.run()
    print(history.final_result())  # should be the proxy's exit IP


asyncio.run(main())

Syntax-checked. ProxySettings and its server, bypass, username and password fields come from Browser Use's source (version 0.13.13 on GitHub); model setup follows the Browser Use docs.

• Browser Use source — ProxySettings (server, bypass, username, password)

Playwright MCP config with a password-protected proxy

Scroll horizontally with the arrow keys to read all content.

{
  "browser": {
    "browserName": "chromium",
    "launchOptions": {
      "proxy": {
        "server": "http://HOST:PORT",
        "username": "USERNAME",
        "password": "PASSWORD"
      }
    }
  }
}

Save as playwright-mcp.json and start the server with npx @playwright/mcp@latest --config playwright-mcp.json. The proxy object is Playwright's launch option, so username and password work here, unlike the --proxy-server flag. JSON validated.

• Playwright MCP — README: --proxy-server and configuration file

Quick fixes

AI agent proxy FAQ

Related setups

Playwright proxies

Playwright proxy setup with username and password: static datacenter IPs for tests and CI, rotating residential from US$1.50/GB when a site blocks servers.

View options and pricing

Static IP proxies for APIs

Proxy for API requests with a static IP: allowlist one fixed datacenter address for cron jobs, CI and scheduled data sync, and keep each job's rate limit to itself.

View options and pricing

Web scraping proxies

Web scraping proxies on real household IPs in 195+ countries and regions: rotating exits or sticky sessions by city, from US$1.50/GB, HTTP and SOCKS5.

View options and pricing