nodemaven-mcp
MCP server that lets AI agents route web requests through NodeMaven residential and mobile proxies, with tools for geo-targeted fetching, proxy checking, and location/usage queries.
README
<img src="docs/img/banner.png" alt="nodemaven-mcp" width="100%">
nodemaven-mcp
English | Русский
Give your AI agent a residential IP. An MCP server that lets Claude Code, Claude Desktop, Cursor or any other MCP client route web requests through NodeMaven residential and mobile proxies: pick an exit country, verify what the target site sees, and pull the page.
you: read this product page the way a shopper in Munich sees it
agent: nodemaven_fetch(url=..., country="de", city="munich")
-> 200, de, 14.99 EUR, exit IP 91.34.x.x (Vodafone), 812 ms
Unofficial, community-built, MIT licensed. Not affiliated with NodeMaven.
Why
Agents are increasingly the thing doing the browsing, and they hit the same wall everyone else does: datacenter IPs get blocked, and prices, stock and search results change by geography. The proxy is already the fix - it just isn't reachable from inside the agent's session.
This server closes that gap in five tools. Nothing here needs a scraping DSL: the agent asks for a country, gets a working connection, and reads the page.
Tools
| Tool | What it does | Needs |
|---|---|---|
nodemaven_build_proxy_url |
Builds a proxy URL with geo targeting encoded in the username, ready for curl, Playwright, Scrapy or requests | proxy creds |
nodemaven_check_proxy |
One request through the proxy: exit IP, geo, ISP, latency | proxy creds |
nodemaven_fetch |
Fetches a URL through the proxy, HTML converted to readable text | proxy creds |
nodemaven_list_locations |
Countries, regions, cities, ISPs, zip codes available for targeting | API key |
nodemaven_get_usage |
Traffic consumed, request counts, per-domain breakdown | API key |
Every tool takes response_format: markdown (default, compact) or json
(complete, for programmatic use), and paginates with limit / offset.
Install
git clone https://github.com/Artirain/nodemaven-mcp
cd nodemaven-mcp
pip install -e .
Requires Python 3.10+.
Configure
Two independent sets of credentials, both from the NodeMaven dashboard:
NODEMAVEN_API_KEY- Profile -> API Key. Used by the location and usage tools.NODEMAVEN_PROXY_USERNAME/NODEMAVEN_PROXY_PASSWORD- Proxy Setup. Used by everything that sends traffic through the proxy.
You can set only one set; tools that need the other say so instead of failing silently.
Claude Code
claude mcp add nodemaven \
-e NODEMAVEN_API_KEY=... \
-e NODEMAVEN_PROXY_USERNAME=... \
-e NODEMAVEN_PROXY_PASSWORD=... \
-- python -m nodemaven_mcp
Claude Desktop / Cursor
claude_desktop_config.json (or .cursor/mcp.json):
{
"mcpServers": {
"nodemaven": {
"command": "python",
"args": ["-m", "nodemaven_mcp"],
"env": {
"NODEMAVEN_API_KEY": "your-api-key",
"NODEMAVEN_PROXY_USERNAME": "your-proxy-username",
"NODEMAVEN_PROXY_PASSWORD": "your-proxy-password"
}
}
}
}
A local .env in the working directory also works - see .env.example.
Environment variables always win over the file.
On Windows, give the absolute path to python.exe rather than bare python.
If Python came from the Microsoft Store, the python on your PATH is an app
execution alias that MCP clients cannot spawn, and the server silently never
appears. python -c "import sys; print(sys.executable)" prints the path to use.
Use it
Three things worth asking your agent, once it is connected:
Check what a site sees
"Am I actually coming out of Germany? Use a Vodafone IP if you can."
The agent calls nodemaven_check_proxy(country="de", isp="vodafone") and reports
the exit IP, city and latency. Useful as a smoke test before a long run.
Read geo-restricted content
"Fetch this page as a US shopper in New York and tell me the price."
nodemaven_fetch(url=..., country="us", city="new york") returns readable text,
truncated to 200 KB by default so it does not flood the context.
Keep one identity across a flow
"Log into the demo account and walk through checkout, same IP the whole time."
Pass the same session_id (4-10 alphanumerics) on every call and NodeMaven pins
the exit IP; drop it and the IP rotates per request.
Targeting spellings matter. nodemaven_list_locations(level="cities", country="de")
gives the exact values NodeMaven accepts, which is the difference between a working
pool and an empty one.
See what it actually does
A tool call is easy to trust and hard to verify. examples/traffic_dashboard.py
runs a local forward proxy with a live dashboard, so you can point the server at
it and watch every hop it makes:
python examples/traffic_dashboard.py # proxy on :8080, dashboard on :8099
NODEMAVEN_PROXY_HOST=127.0.0.1
NODEMAVEN_HTTP_PORT=8080

Each row is a real network request. The route column is decoded from the proxy
username, which is exactly the string NodeMaven uses to choose the exit IP - so you
can confirm that country="de", city="Munich" in a tool call really did leave your
machine as DE / MUNICH, and that a session_id stayed pinned across calls.
The dashboard is a development aid: it accepts any credentials and does not encrypt anything. Do not put it on a network you share.
Design notes
Passwords stay masked. build_proxy_url returns *** in place of the
password unless the caller passes reveal_credentials=true. Transcripts get
shared, pasted into issues and fed back into models; a proxy password should not
ride along by default.
Errors are written for an agent, not a log file. A refused connection says that an over-narrow city or ISP filter is the usual cause and suggests widening it. A 401 names the environment variable to check. An agent that can read the error can retry correctly on the first attempt.
Targeting is pure and tested. NodeMaven encodes location in the proxy username:
acct-country-us-region-california-city-los_angeles-isp-t_mobile_usa-sid-ab12cd-filter-medium
Building that string - normalizing "Los Angeles" to los_angeles, rejecting a
session id that would silently disable stickiness, validating the port against
the range for the chosen protocol - is pure logic, so it is covered by tests that
never touch the network.
No parser dependency. HTML-to-text runs on html.parser from the standard
library. The dependency list is mcp, httpx and pydantic.
Development
pip install -e ".[dev]"
pytest -q # 60 tests, no network, no credentials needed
ruff check .
CI runs the suite on Python 3.10, 3.11 and 3.12.
Contributions welcome - see CONTRIBUTING.md.
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.