AgentProxy

Agent-to-agent proxy MCP server — fetch any URL through 2M+ residential IPs, bypass anti-bot systems, render JS-heavy pages, geo-target by country or city, maintain sticky sessions.

Powered by Novada — sign up once, all tools from one account.

Get Your Credentials

1. Sign up at novada.com — 30 seconds, no credit card
2. From your dashboard, get the credentials for each tool you need:

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 don't need all of them — only set what you use.

Install

Search only:

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

Fetch + Session (full proxy network):

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

All tools:

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

NOVADA_PROXY_HOST is your account-specific proxy host from the Endpoint Generator (e.g. abc123.vtv.na.novada.pro). Required for reliable sticky sessions.

Why AgentProxy

AI agents get blocked on 60–70% of commercial websites with standard HTTP requests. AgentProxy routes through real home devices — Android phones, Windows PCs, Macs — so your agent looks like a real user.

Problem	AgentProxy
Amazon, LinkedIn block your agent	Residential IPs from real devices
Cloudflare / Akamai challenges	Bypassed via real device fingerprints
JS-rendered pages show blank	Browser API runs real Chromium
Geo-restricted content	195+ countries, city-level targeting
Multi-step workflows need same IP	Sticky sessions — same IP across calls
Need structured search results	Built-in Google search via Novada

Tools

agentproxy_fetch

Fetch any URL through Novada's residential proxy. Works on Amazon, LinkedIn, Cloudflare-protected pages, and most commercial sites.

Requires: NOVADA_PROXY_USER + NOVADA_PROXY_PASS

url        — Target URL (required, http/https only)
country    — 2-letter code: US, DE, JP, GB, BR, ... (195+ countries)
city       — City-level: newyork, london, tokyo, ...
session_id — Reuse same ID to keep the same IP (letters/numbers/underscores, no hyphens)
format     — "markdown" (default) | "raw"
timeout    — Seconds, 1-120 (default 60)

agentproxy_render [BETA]

Render JavaScript-heavy pages using Novada's Browser API (real Chromium). Use this for SPAs, React/Vue apps, and sites that need JS to load content.

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

url       — Target URL (required)
format    — "markdown" (default) | "html" | "text"
wait_for  — CSS selector to wait for before extracting (e.g. ".product-title")
timeout   — Seconds, 5-120 (default 60)

agentproxy_search

Structured web search via Novada. Returns clean titles, URLs, descriptions — no HTML parsing needed.

Requires: NOVADA_API_KEY

query    — Search query (required)
num      — Results count, 1-20 (default 10)
country  — Localize results (e.g. us, uk, de)
language — Language code (e.g. en, zh, de)

agentproxy_session

Sticky session fetch — every call with the same session_id uses the same residential IP. For login flows, paginated scraping, price monitoring.

Requires: NOVADA_PROXY_USER + NOVADA_PROXY_PASS

session_id — Unique session ID, no hyphens (required)
url        — Target URL (required)
country    — 2-letter country code
format     — "markdown" | "raw"
timeout    — Seconds, 1-120

agentproxy_status

Check Novada's proxy network health — node count, device breakdown, service status.

Example: Agent Price Monitor

# Check Amazon price from US residential IP
agentproxy_fetch(url="https://amazon.com/dp/B0BSHF7WHW", country="US")

# Same product, three markets
agentproxy_fetch(url=..., country="US")
agentproxy_fetch(url=..., country="DE")
agentproxy_fetch(url=..., country="JP")

# Render a JS-heavy real estate page
agentproxy_render(url="https://zillow.com/homes/NYC_rb/", wait_for=".list-card")

# Multi-page scrape with same IP (no hyphens in session ID)
agentproxy_session(session_id="job001page1", url="https://example.com/page/1")
agentproxy_session(session_id="job001page1", url="https://example.com/page/2")

# Web search
agentproxy_search(query="AI proxy tools 2025", num=5)

Network

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

Confirmed Working

Amazon, LinkedIn, Cloudflare-protected sites, HackerNews, GitHub, Wikipedia, BBC, CNN, Reddit, IMDB, and more.

Known Limitations

• Sites requiring full JS execution → use agentproxy_render
• agentproxy_render requires a separate Browser API subscription and NOVADA_BROWSER_WS env var
• Session IDs must not contain hyphens
• For reliable sticky sessions, set NOVADA_PROXY_HOST to your account-specific proxy host

License

MIT — Powered by Novada