Proxy4Agent

Residential proxy MCP server for AI agents. Route any HTTP request through 2M+ real home devices — Android phones, Windows PCs, Macs — to bypass anti-bot systems, geo-target by country or city, and maintain sticky sessions across multi-step workflows.

Works with Claude Code, Cursor, Windsurf, Cline, Continue, and any MCP-compatible AI agent. Powered by Novada.

---

Why Proxy4Agent

AI agents get blocked on 60–70% of commercial websites. Standard HTTP requests are detected and rejected by Cloudflare, Akamai, DataDome, PerimeterX, and similar systems. Proxy4Agent routes your agent through real residential IPs — so it looks indistinguishable from a human browser.

Problem	Proxy4Agent
Amazon, LinkedIn block your agent	Residential IPs from real home devices
Cloudflare / Akamai bot challenges	Bypassed via real device fingerprints
JS-rendered pages return blank	Browser API runs real Chromium
Geo-restricted or localized content	195+ countries, city-level targeting
Multi-step workflows lose session	Sticky sessions — same IP across calls
Need structured search results	Built-in Google search, clean JSON output

---

Providers

Proxy4Agent works with any HTTP proxy. Novada is the built-in default with the deepest integration. BrightData, Smartproxy, and Oxylabs have dedicated adapters with full auto-targeting. Any other provider works via the generic adapter.

Priority: Novada → BrightData → Smartproxy → Oxylabs → Generic. First configured provider wins.

Novada (default — recommended)

Full integration: automatic geo-targeting, city-level targeting, sticky sessions, 195+ countries.

claude mcp add bestproxy4agents \
  -e NOVADA_PROXY_USER=your_username \
  -e NOVADA_PROXY_PASS=your_password \
  -- npx -y bestproxy4agents

Get credentials: novada.com → Dashboard → Residential Proxies → Endpoint Generator

BrightData

Full integration: automatic country/city/session targeting via BrightData's username-suffix format.

claude mcp add bestproxy4agents \
  -e BRIGHTDATA_USER="brd-customer-abc123-zone-residential" \
  -e BRIGHTDATA_PASS=your_password \
  -- npx -y bestproxy4agents

Get credentials: brightdata.com → Proxies & Scraping → Residential → Access Parameters

BRIGHTDATA_USER is your full username including zone (e.g. brd-customer-abc123-zone-residential). Optional: BRIGHTDATA_HOST (default: zproxy.lum-superproxy.io), BRIGHTDATA_PORT (default: 22225).

Smartproxy

Full integration: automatic country/city/session targeting.

claude mcp add bestproxy4agents \
  -e SMARTPROXY_USER=your_username \
  -e SMARTPROXY_PASS=your_password \
  -- npx -y bestproxy4agents

Get credentials: smartproxy.com → Dashboard → Residential → Endpoint Generator

Optional: SMARTPROXY_HOST (default: gate.smartproxy.com), SMARTPROXY_PORT (default: 10001).

Oxylabs

Full integration: automatic country/city/session targeting.

claude mcp add bestproxy4agents \
  -e OXYLABS_USER=your_username \
  -e OXYLABS_PASS=your_password \
  -- npx -y bestproxy4agents

Get credentials: oxylabs.io → Dashboard → Residential Proxies → Access Details

Optional: OXYLABS_HOST (default: pr.oxylabs.io), OXYLABS_PORT (default: 7777).

Generic HTTP Proxy — any other provider

Set PROXY_URL to use IPRoyal, your own infrastructure, or any standard HTTP proxy.

# IPRoyal
claude mcp add bestproxy4agents \
  -e PROXY_URL="http://username:password@geo.iproyal.com:12321" \
  -- npx -y bestproxy4agents

# Any HTTP proxy
claude mcp add bestproxy4agents \
  -e PROXY_URL="http://user:pass@your-proxy-host:port" \
  -- npx -y bestproxy4agents

Note: With the generic adapter, encode country/city/session targeting directly in your proxy URL per your provider's format. country, city, and session_id tool parameters are logged as warnings and not forwarded.

Feature	Novada	BrightData	Smartproxy	Oxylabs	Generic HTTP
Auto country targeting	✓	✓	✓	✓	manual
Auto city targeting	✓	✓	✓	✓	manual
Sticky sessions	✓	✓	✓	✓	manual
Built-in search API	✓	—	—	—	—
Browser API (JS render)	✓	—	—	—	—

---

Get Your Credentials

Sign up at novada.com — 30 seconds, no credit card.

Tool	Required env vars	Where to get them
agentproxy_fetch	NOVADA_PROXY_USER + NOVADA_PROXY_PASS	Dashboard → Residential Proxies → Endpoint Generator
agentproxy_session	NOVADA_PROXY_USER + NOVADA_PROXY_PASS	Dashboard → Residential Proxies → Endpoint Generator
agentproxy_search	NOVADA_API_KEY	Dashboard → API Keys
agentproxy_render [BETA]	NOVADA_BROWSER_WS	Dashboard → Browser API → Playground → copy Puppeteer URL
agentproxy_status	(none)	—

You only need credentials for the tools you use.

---

Install

Fetch + Session (core — recommended start):

claude mcp add bestproxy4agents \
  -e NOVADA_PROXY_USER=your_username \
  -e NOVADA_PROXY_PASS=your_password \
  -e NOVADA_PROXY_HOST=your_account_host \
  -- npx -y bestproxy4agents

Search only:

claude mcp add bestproxy4agents \
  -e NOVADA_API_KEY=your_key \
  -- npx -y bestproxy4agents

All tools:

claude mcp add bestproxy4agents \
  -e NOVADA_PROXY_USER=your_username \
  -e NOVADA_PROXY_PASS=your_password \
  -e NOVADA_PROXY_HOST=your_account_host \
  -e NOVADA_API_KEY=your_key \
  -- npx -y bestproxy4agents

NOVADA_PROXY_HOST — your account-specific proxy host from the Endpoint Generator (e.g. abc123.vtv.na.novada.pro). Required for reliable sticky sessions. Defaults to the shared load balancer if omitted.

---

Compatible With

Proxy4Agent works with any MCP-compatible AI client:

Client	Install method
Claude Code	claude mcp add bestproxy4agents -e ... -- npx -y bestproxy4agents
Cursor	Settings → MCP → Add server → npx -y bestproxy4agents
Windsurf	MCP config → npx -y bestproxy4agents
Cline	MCP settings → command: npx, args: ["-y", "bestproxy4agents"]
Continue	.continue/config.json → mcpServers
Smithery	smithery.ai/server/proxy4agent
Any MCP client	stdio transport, npx -y bestproxy4agents

Claude Code example (copy-paste ready):

claude mcp add bestproxy4agents \
  -e NOVADA_PROXY_USER=your_username \
  -e NOVADA_PROXY_PASS=your_password \
  -- npx -y bestproxy4agents

Cursor / Windsurf / Cline — add to your MCP config:

{
  "mcpServers": {
    "bestproxy4agents": {
      "command": "npx",
      "args": ["-y", "bestproxy4agents"],
      "env": {
        "NOVADA_PROXY_USER": "your_username",
        "NOVADA_PROXY_PASS": "your_password"
      }
    }
  }
}

---

Tools

agentproxy_fetch

Fetch any URL through a residential proxy. Works on Amazon, LinkedIn, Cloudflare-protected pages, and most anti-bot-protected sites.

Requires: Any proxy provider — Novada (NOVADA_PROXY_USER + NOVADA_PROXY_PASS), BrightData, Smartproxy, Oxylabs, or PROXY_URL

Parameter	Type	Default	Description
url	string	required	Target URL (http:// or https://)
country	string	—	2-letter ISO code: US, DE, JP, GB, BR, IN, FR, CA, AU... (195+ countries)
city	string	—	City-level targeting: newyork, london, tokyo, paris, berlin...
session_id	string	—	Reuse same ID to keep the same IP (letters/numbers/underscores only, max 64 chars)
format	string	markdown	markdown strips HTML tags · raw returns full HTML
timeout	number	60	Timeout in seconds (1–120)

---

agentproxy_session

Sticky session fetch — every call with the same session_id uses the same residential IP. Essential for login flows, paginated scraping, and price monitoring across pages.

Requires: Any proxy provider — Novada (NOVADA_PROXY_USER + NOVADA_PROXY_PASS), BrightData, Smartproxy, Oxylabs, or PROXY_URL

Parameter	Type	Default	Description
session_id	string	required	Unique ID — reuse to keep same IP (no hyphens, max 64 chars)
url	string	required	Target URL
country	string	—	2-letter country code
city	string	—	City-level targeting: newyork, london, tokyo...
format	string	markdown	markdown or raw
timeout	number	60	Timeout in seconds (1–120)

---

agentproxy_search

Structured Google search via Novada. Returns titles, URLs, and descriptions — no HTML parsing needed. Best for discovery and research tasks.

Requires: NOVADA_API_KEY

Parameter	Type	Default	Description
query	string	required	Search query
num	number	10	Result count (1–20)
country	string	—	Localize results: us, uk, de, jp, fr...
language	string	—	Language: en, zh, de, ja, fr...

---

agentproxy_render [BETA]

Render JavaScript-heavy pages using Novada's Browser API (real Chromium, full JS execution). Use for SPAs, React/Vue apps, and pages that return blank without a real browser.

Requires: NOVADA_BROWSER_WS (copy the Puppeteer URL from Dashboard → Browser API → Playground)

Parameter	Type	Default	Description
url	string	required	Target URL
format	string	markdown	markdown · html · text
wait_for	string	—	CSS selector to wait for before extracting (e.g. .product-title)
timeout	number	60	Timeout in seconds (5–120)

---

agentproxy_status

Check Novada's proxy network health — live node count, device types, service status. No credentials required.

---

Real-World Results

Live outputs from actual API calls — not fabricated.

Geo-targeting: same URL, different exit countries

agentproxy_fetch(url="https://httpbin.org/ip", country="US", format="raw")
→ { "origin": "200.50.235.236" }   ← US residential IP

agentproxy_fetch(url="https://httpbin.org/ip", country="JP", format="raw")
→ { "origin": "60.85.57.175" }     ← Japan residential IP

Sticky session: same IP confirmed across two requests

agentproxy_session(session_id="job001", url="https://httpbin.org/ip", format="raw")
→ { "origin": "103.135.135.168" }

agentproxy_session(session_id="job001", url="https://httpbin.org/ip", format="raw")
→ { "origin": "103.135.135.168" }  ← same IP, confirmed ✓

Amazon — 1.6 MB product page, not blocked

agentproxy_fetch(url="https://www.amazon.com/dp/B0BSHF7WHW", country="US")
→ [URL: https://www.amazon.com/dp/B0BSHF7WHW | Status: 200 | Size: 1637 KB | Country: US]

  Apple 2023 MacBook Pro — M2 Pro chip, 16-inch, 16GB, 1TB
  Full product page: price, reviews, specs, related items

HackerNews — 30 stories as clean markdown

agentproxy_fetch(url="https://news.ycombinator.com")
→ [URL: https://news.ycombinator.com | Status: 200 | Size: 34 KB]

  1. LittleSnitch for Linux — 752 points, 243 comments
  2. I ported Mac OS X to the Nintendo Wii — 1590 points, 281 comments
  3. Git commands I run before reading any code — 2054 points, 445 comments
  ...

Google search — structured output, no HTML parsing

agentproxy_search(query="residential proxy for AI agents", num=3)
→ Search: "residential proxy for AI agents" via GOOGLE — 3 results

  1. Residential Proxies Trusted by Fortune 500 Companies
     https://brightdata.com/proxy-types/residential-proxies
     Access 400M+ residential proxies from 195 countries...

  2. Proxies for AI Web Agents: The Complete Guide
     https://netnut.io/proxies-for-ai-web-agents/
     Real-user IPs that bypass even the toughest anti-bot systems...

Browser render — React SPA, full JS execution

agentproxy_render(url="https://react.dev", format="markdown")
→ [URL: https://react.dev | Title: React | Size: 266 KB | Rendered: yes (Browser API)]

  React v19.2

  The library for web and native user interfaces

  ## Create user interfaces from components
  React lets you build user interfaces out of individual pieces called components.
  Create your own React components like Thumbnail, LikeButton, and Video.
  Then combine them into entire screens, pages, and apps.
  ...

Browser render — wait_for selector on HackerNews

agentproxy_render(url="https://news.ycombinator.com", wait_for=".athing")
→ [URL: https://news.ycombinator.com | Title: Hacker News | Size: 34 KB | Rendered: yes (Browser API)]

  1. LittleSnitch for Linux — 834 points, 285 comments
  2. Help Keep Thunderbird Alive — 138 points, 77 comments
  3. I ported Mac OS X to the Nintendo Wii — 1638 points
  ...

---

Example Workflows

Price monitor — same product, three markets

agentproxy_fetch(url="https://amazon.com/dp/B0BSHF7WHW", country="US")
agentproxy_fetch(url="https://amazon.com/dp/B0BSHF7WHW", country="DE")
agentproxy_fetch(url="https://amazon.com/dp/B0BSHF7WHW", country="JP")

Login + multi-page scrape with same IP

agentproxy_session(session_id="workflow01", url="https://example.com/login")
agentproxy_session(session_id="workflow01", url="https://example.com/dashboard")
agentproxy_session(session_id="workflow01", url="https://example.com/data/page/1")
agentproxy_session(session_id="workflow01", url="https://example.com/data/page/2")

Research pipeline

# 1. Find relevant pages
agentproxy_search(query="Claude MCP proxy tools", num=10)

# 2. Fetch each result through residential proxy
agentproxy_fetch(url="https://found-result.com/article", country="US")

# 3. Render JS-heavy dashboard (requires Browser API)
agentproxy_render(url="https://app.example.com/dashboard", wait_for=".data-table")

---

Geo Coverage

195+ countries including:

US GB DE FR JP CA AU BR IN KR SG NL IT ES MX RU PL SE NO DK FI CH AT BE PT CZ HU RO UA TR IL ZA NG EG AR CL CO PE VN TH ID MY PH PK BD TW HK NZ + 188 more

City-level targeting (selected): newyork · losangeles · chicago · london · paris · berlin · tokyo · seoul · sydney · toronto · singapore · dubai · mumbai · saopaulo

---

Network

Metric	Value
Residential IPs	2,000,000+
Live nodes	7,000+
Countries	195+
Device types	Android, Windows, Mac
Uptime	99.9%

---

Confirmed Working

E-commerce: Amazon, eBay, Walmart, Etsy, Shopify stores
Professional networks: LinkedIn
Anti-bot protected: Cloudflare sites, Akamai-protected pages, DataDome-protected sites
News & content: HackerNews, Reddit, BBC, CNN, NYTimes
Tech: GitHub, Wikipedia, Stack Overflow
Entertainment: IMDB, Rotten Tomatoes

---

Dependencies

Package	Version	Purpose
@modelcontextprotocol/sdk	^1.26	MCP protocol — stdio transport, tool definitions, request handling
axios	^1.7	HTTP client — handles redirects, compression, streaming
https-proxy-agent	^9.0	HTTPS proxy routing — CONNECT tunnel + TLS for secure targets
http-proxy-agent	^7.0	HTTP proxy routing — plain HTTP targets through proxy
puppeteer-core	^22.15	Browser API — WebSocket connection to Novada's real Chromium cloud

Lightweight core. Package size: ~52 KB (excluding node_modules). puppeteer-core is the largest dependency but bundles no browser — it connects to Novada's remote Chromium.

---

Use Cases

For AI agents that need to:

• Scrape e-commerce sites (Amazon, eBay, Walmart, Shopify) without getting blocked
• Monitor prices across different countries and markets
• Access geo-restricted content from 195+ countries
• Perform competitive intelligence on Cloudflare/Akamai-protected sites
• Execute multi-step login flows with consistent IP (sticky sessions)
• Render JavaScript-heavy SPAs that return blank without a real browser
• Search Google programmatically with clean structured results
• Collect localized search results from different regions
• Access professional networks (LinkedIn) behind anti-bot systems

For developers building:

• AI-powered web research tools
• Price comparison agents
• Content aggregation pipelines
• SEO monitoring dashboards
• Market research automation
• Lead generation workflows

---

Feedback & Support

We'd love to hear from you — feature requests, bug reports, or just how you're using proxy4agent:

• Email: tong.wu@novada.com
• GitHub Issues: github.com/Goldentrii/proxy4agent/issues
• Website: novada.com

Your feedback drives our roadmap. Every bug report and feature request helps make proxy4agent better for all agents and developers.

---

Known Limitations

• Sites requiring full JS execution → use agentproxy_render
• agentproxy_render requires a separate Novada Browser API subscription and NOVADA_BROWSER_WS
• Session IDs must not contain hyphens (Novada uses - as its auth delimiter)
• For reliable sticky sessions, set NOVADA_PROXY_HOST to your account-specific host

---

License

MIT © Novada

Permission is hereby granted, free of charge, to any person obtaining a copy of this software to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies, subject to the following conditions: the above copyright notice and this permission notice shall be included in all copies or substantial portions of the software.

The software is provided "as is", without warranty of any kind.

See LICENSE for full text.

---

---

Proxy4Agent（中文文档）

AI 智能体的住宅代理 MCP 服务器。 通过 200 万+ 真实家庭设备路由 HTTP 请求，绕过反机器人系统，按国家/城市定位，跨请求保持同一 IP。

支持 Claude Code、Cursor、Windsurf、Cline、Continue 及所有 MCP 兼容智能体。由 Novada 提供支持。

---

为什么选择 Proxy4Agent

AI 智能体在 60-70% 的商业网站上被封锁。标准 HTTP 请求会被 Cloudflare、Akamai、DataDome 等系统检测并拒绝。Proxy4Agent 通过真实住宅 IP 路由请求 — 与人类浏览器无法区分。

问题	Proxy4Agent
Amazon、LinkedIn 封锁你的智能体	来自真实家庭设备的住宅 IP
Cloudflare / Akamai 机器人检测	通过真实设备指纹绕过
JS 渲染页面返回空白	浏览器 API 运行真实 Chromium
地区限制或本地化内容	195+ 国家，城市级定位
多步骤工作流丢失会话	粘性会话 — 同一 IP 跨请求
需要结构化搜索结果	内置 Google 搜索，输出 JSON

---

支持的代理提供商

优先级：Novada → BrightData → Smartproxy → Oxylabs → 通用 HTTP。首个配置的提供商生效。

功能	Novada	BrightData	Smartproxy	Oxylabs	通用 HTTP
自动国家定位	✓	✓	✓	✓	手动
自动城市定位	✓	✓	✓	✓	手动
粘性会话	✓	✓	✓	✓	手动
搜索 API	✓	—	—	—	—
浏览器 API (JS 渲染)	✓	—	—	—	—

---

快速开始

# Novada（推荐）
claude mcp add bestproxy4agents \
  -e NOVADA_PROXY_USER=你的用户名 \
  -e NOVADA_PROXY_PASS=你的密码 \
  -- npx -y bestproxy4agents

# BrightData
claude mcp add bestproxy4agents \
  -e BRIGHTDATA_USER="brd-customer-abc123-zone-residential" \
  -e BRIGHTDATA_PASS=你的密码 \
  -- npx -y bestproxy4agents

# 任意 HTTP 代理
claude mcp add bestproxy4agents \
  -e PROXY_URL="http://user:pass@host:port" \
  -- npx -y bestproxy4agents

获取 Novada 凭证：novada.com → 仪表盘 → 住宅代理 → 端点生成器

---

5 个 MCP 工具

工具	功能
agentproxy_fetch	通过住宅代理获取任何 URL。支持国家/城市/会话定位。
agentproxy_session	粘性会话 — 同一 session_id 使用同一 IP。用于登录流程、分页抓取。
agentproxy_search	Google 搜索，返回标题、URL、描述。无需 HTML 解析。
agentproxy_render	用真实 Chromium 渲染 JS 页面（SPA、React/Vue 应用）。
agentproxy_status	检查代理网络健康状态。无需凭证。

---

使用场景

AI 智能体需要：

• 抓取电商网站（Amazon、eBay、Walmart、Shopify）不被封锁
• 跨国家监控价格
• 访问 195+ 国家的地区限制内容
• 对 Cloudflare/Akamai 保护的站点进行竞争情报收集
• 执行多步登录流程并保持同一 IP
• 渲染无真实浏览器就返回空白的 JS 页面
• 获取不同地区的本地化搜索结果

开发者构建：

• AI 驱动的网页研究工具
• 价格比较智能体
• 内容聚合管道
• SEO 监控仪表盘
• 市场调研自动化

---

反馈与支持

• 邮箱： tong.wu@novada.com
• GitHub Issues： github.com/Goldentrii/proxy4agent/issues
• 网站： novada.com

---

许可证

MIT © Novada