Aluvia Python SDK

Stop getting blocked. Aluvia routes your AI agent's web traffic through premium US mobile carrier IPs — the same IPs used by real people on their phones. Websites trust them, so your agent stops hitting 403s, CAPTCHAs, and rate limits.

This SDK gives you everything you need:

• CLI for browser automation — launch headless Chromium sessions from the command line, with JSON output designed for AI agent frameworks
• Automatic block detection and unblocking — the SDK detects 403s, WAF challenges, and CAPTCHAs, then reroutes through Aluvia and reloads the page automatically
• Smart routing — proxy only the sites that block you; everything else goes direct to save cost and latency
• Runtime rule updates — add hostnames to proxy rules on the fly, no restarts or redeployments
• Adapters for popular tools — Playwright, Selenium, httpx, requests, and aiohttp
• IP rotation and geo targeting — rotate IPs or target specific US regions at runtime
• REST API wrapper — manage connections, check usage, and build custom tooling with AluviaApi
• MCP server — for Model Context Protocol (MCP) only, use the separate package: pip install aluvia-mcp and run aluvia-mcp. See aluvia_mcp/README.md for details.

Table of Contents

• Quick Start
• MCP Server (Model Context Protocol)
• Aluvia Client
• Architecture
• Operating Modes
• Using Aluvia Client
• Routing Rules
• Dynamic Unblocking
• Tool Integration Adapters
• Aluvia API
• License

Quick Start

1. Get Aluvia API key

Aluvia dashboard

2. Install

pip install aluvia-sdk playwright
export ALUVIA_API_KEY="your-api-key"

3. Run

Aluvia automatically detects website blocks and uses mobile IPs when necessary.

from aluvia_sdk import AluviaClient
import asyncio

async def main():
    client = AluviaClient(api_key="your-api-key", start_playwright=True)
    connection = await client.start()
    
    page = await connection.browser.new_page()
    await page.goto("https://example.com")
    print(await page.title())
    
    await connection.close()

asyncio.run(main())

MCP Server (Model Context Protocol)

For AI agent frameworks that support MCP (Claude Desktop, Claude Code, Cursor, etc.), use the Aluvia MCP server:

pip install aluvia-mcp
export ALUVIA_API_KEY="your-api-key"
aluvia-mcp

The MCP server exposes all Aluvia CLI functionality as structured tools. See aluvia_mcp/README.md for configuration and usage.

Aluvia client

The Aluvia client runs a local rules-based proxy server on your agent's host, handles authentication and connection management, and provides ready-to-use adapters for popular tools like Playwright, Selenium, and httpx.

Simply point your automation tool at the local proxy address (127.0.0.1) and the client handles the rest. For each request, the client checks the destination hostname against user-defined (or agent-defined) routing rules and decides whether to send it through Aluvia's mobile IPs or direct to the destination.

┌──────────────────┐      ┌──────────────────────────┐      ┌──────────────────────┐
│                  │      │                          │      │                      │
│    Your Agent    │─────▶     Aluvia Client         ─────▶  gateway.aluvia.io    │
│                  │      │     127.0.0.1:port       │      │    (Mobile IPs)      │
│                  │      │                          │      │                      │
└──────────────────┘      │  Per-request routing:    │      └──────────────────────┘
                          │                          │
                          │  not-blocked.com ──────────────▶ Direct
                          │  blocked-site.com ─────────────▶ Via Aluvia
                          │                          │
                          └──────────────────────────┘

Benefits:

• Avoid blocks: Websites flag datacenter IPs as bot traffic, leading to 403s, CAPTCHAs, and rate limits. Mobile IPs appear as real users, so requests go through.
• Reduce costs and latency: Hostname-based routing rules let you proxy only the sites that need it. Traffic to non-blocked sites goes direct, saving money and reducing latency.
• Unblock without restarts: Rules update at runtime. When a site blocks your agent, add it to the proxy rules and retry—no need to restart workers or redeploy.
• Simplify integration: One SDK with ready-to-use adapters for Playwright, Selenium, httpx, requests, and aiohttp.

Quick start

Understand the basics

• What is Aluvia?
• Understanding connections

Get Aluvia API key

1. Create an account at dashboard.aluvia.io
2. Go to API and SDKs and get your API Key

Install the SDK

pip install aluvia-sdk

Requirements: Python 3.9 or later

Example: Dynamic unblocking with Playwright

This example shows how an agent can use the Aluvia client to dynamically unblock websites. It demonstrates starting the client, using the Playwright integration adapter, configuring geo targeting and session ID, detecting blocks, and updating routing rules on the fly.

import asyncio
from playwright.async_api import async_playwright
from aluvia_sdk import AluviaClient

async def main():
    # Initialize the Aluvia client with your API key
    client = AluviaClient(api_key="your-api-key")

    # Start the client (launches local proxy, fetches connection config)
    connection = await client.start()

    # Configure geo targeting (use California IPs)
    await client.update_target_geo("us_ca")

    # Set session ID (requests with the same session ID use the same IP)
    await client.update_session_id("agentsession1")

    # Launch browser using the Playwright integration adapter
    # The adapter returns proxy settings in Playwright's expected format
    async with async_playwright() as p:
        browser = await p.chromium.launch(proxy=connection.as_playwright())

        # Track hostnames we've added to proxy rules
        proxied_hosts = set()

        async def visit_with_retry(url: str) -> str:
            page = await browser.new_page()
            try:
                response = await page.goto(url, wait_until="domcontentloaded")
                hostname = url.split("//")[1].split("/")[0]

                # Detect if the site blocked us (403, 429, or challenge page)
                status = response.status if response else 0
                title = await page.title()
                is_blocked = status in (403, 429) or "blocked" in title.lower()

                if is_blocked and hostname not in proxied_hosts:
                    print(f"Blocked by {hostname} — adding to proxy rules")

                    # Update routing rules to proxy this hostname through Aluvia
                    # Rules update at runtime—no need to restart the browser
                    proxied_hosts.add(hostname)
                    await client.update_rules(list(proxied_hosts))

                    # Rotate to a fresh IP by changing the session ID
                    import time
                    await client.update_session_id(f"retry{int(time.time())}")

                    await page.close()
                    return await visit_with_retry(url)

                return await page.content()
            finally:
                await page.close()

        try:
            # First attempt goes direct; if blocked, retries through Aluvia
            html = await visit_with_retry("https://example.com/data")
            print("Success:", html[:200])
        finally:
            # Always close the browser and connection when done
            await browser.close()
            await connection.close()

if __name__ == '__main__':
    asyncio.run(main())

Example: Auto-launch Playwright browser

For even simpler setup, the SDK can automatically launch a Chromium browser that's already configured with the Aluvia proxy. This eliminates the need to manually import Playwright and configure proxy settings.

import asyncio
from aluvia_sdk import AluviaClient

async def main():
    # Initialize with start_playwright option to auto-launch browser
    client = AluviaClient(
        api_key="your-api-key",
        start_playwright=True,  # Automatically launch and configure Chromium
    )

    # Start the client - this also launches the browser
    connection = await client.start()

    # Browser is already configured with Aluvia proxy
    browser = connection.browser
    page = await browser.new_page()

    # Configure geo targeting and session ID
    await client.update_target_geo("us_ca")
    await client.update_session_id("session1")

    # Navigate directly - proxy is already configured
    await page.goto("https://example.com")
    print("Title:", await page.title())

    # Cleanup - automatically closes both browser and proxy
    await connection.close()

if __name__ == '__main__':
    asyncio.run(main())

Note: To use start_playwright=True, you must install Playwright:

pip install playwright
playwright install chromium

Integration guides

The Aluvia client provides ready-to-use adapters for popular automation and HTTP tools. Check the integration examples in the Node.js SDK docs for reference patterns that can be adapted to Python.

Architecture

The client is split into two independent planes:

┌─────────────────────────────────────────────────────────────────┐
│                        AluviaClient                             │
├─────────────────────────────┬───────────────────────────────────┤
│       Control Plane         │          Data Plane               │
│       (ConfigManager)       │          (ProxyServer)            │
├─────────────────────────────┼───────────────────────────────────┤
│ • Fetches/creates config    │ • Local HTTP proxy (proxy.py)     │
│ • Polls for updates (ETag)  │ • Per-request routing decisions   │
│ • PATCH updates (rules,     │ • Uses rules engine to decide:    │
│   session, geo)             │   direct vs gateway               │
└─────────────────────────────┴───────────────────────────────────┘

Control Plane (ConfigManager)

• Communicates with the Aluvia REST API (/account/connections/...)
• Fetches proxy credentials and routing rules
• Polls for configuration updates
• Pushes updates (rules, session ID, geo)

Data Plane (ProxyServer)

• Runs a local HTTP proxy on 127.0.0.1
• For each request, uses the rules engine to decide whether to route direct or via Aluvia.
• Because the proxy reads the latest config per-request, rule updates take effect immediately

Operating modes

The Aluvia client has two operating modes: Client Proxy Mode (default) and Gateway Mode.

Client Proxy Mode

How it works: The SDK runs a local proxy on 127.0.0.1. For each request, it checks your routing rules and sends traffic either direct or through Aluvia.

Why use it:

• Selective routing reduces cost and latency (only proxy what you need)
• Credentials stay inside the SDK (nothing secret in your config)
• Rule changes apply immediately (no restarts)

Best for: Using per-hostname routing rules.

Gateway Mode

Set local_proxy=False to enable.

How it works: No local proxy. Your tools connect directly to gateway.aluvia.io and ALL traffic goes through Aluvia.

Why use it:

• No local process to manage
• Simpler setup for tools with native proxy auth support

Best for: When you want all traffic proxied without selective routing.

Using Aluvia client

1. Create a client

client = AluviaClient(
    api_key=os.environ["ALUVIA_API_KEY"],
    connection_id=123,       # Optional: reuse an existing connection
    local_proxy=True,        # Optional: default True (recommended)
    start_playwright=True,   # Optional: auto-launch Chromium browser
)

2. Start the client and get a connection

connection = await client.start()

This starts the local proxy and returns a connection object you'll use with your tools. Understanding the connection object

3. Use the connection with your tools

Pass the connection to your automation tool using the appropriate adapter:

browser = await p.chromium.launch(proxy=connection.as_playwright())

4. Update routing as necessary

While your agent is running, you can update routing rules, rotate IPs, or change geo targeting—no restart needed:

await client.update_rules(["blocked-site.com"])    # Add hostname to proxy rules
await client.update_session_id("newsession")       # Rotate to a new IP
await client.update_target_geo("us_ca")            # Target California IPs

5. Clean up when done

await connection.close()  # Stops proxy, polling, and releases resources

Routing rules

The Aluvia Client starts a local proxy server that routes each request based on hostname rules that you (or our agent) set. Rules can be updated at runtime without restarting the agent.

Traffic can be sent either:

• direct (using the agent's datacenter/cloud IP) or,
• through Aluvia's mobile proxy IPs,

Benefits

• Selectively routing traffic to mobile proxies reduces proxy costs and connection latency.
• Rules can be updated during runtime, allowing agents to work around website blocks on the fly.

Example rules

await client.update_rules(["*"])                              # Proxy all traffic
await client.update_rules(["target-site.com", "*.google.com"]) # Proxy specific hosts
await client.update_rules(["*", "-api.stripe.com"])           # Proxy all except specified
await client.update_rules([])                                 # Route all traffic direct

Supported routing rule patterns:

Dynamic unblocking

Most proxy solutions require you to decide upfront which sites to proxy. If a site blocks you later, you're stuck—restart your workers, redeploy your fleet, or lose the workflow.

With Aluvia, your agent can unblock itself. When a request fails with a 403 or 429, your agent adds that hostname to its routing rules and retries. The update takes effect immediately—no restart, no redeployment, no lost state.

This turns blocking from a workflow-ending failure into a minor speed bump.

response = await page.goto(url)

if response and response.status in (403, 429):
    # Blocked! Add this hostname to proxy rules and retry
    hostname = url.split("//")[1].split("/")[0]
    await client.update_rules([*current_rules, hostname])
    await page.goto(url)  # This request goes through Aluvia

Your agent learns which sites need proxying as it runs. Sites that don't block you stay direct (faster, cheaper). Sites that do block you get routed through mobile IPs automatically.

Tool integration adapters

Every tool has its own way of configuring proxies—Playwright wants a dict with server/username/password, Selenium wants a string, httpx wants an agent, and some tools don't support proxies at all. The SDK handles all of this for you:

Playwright auto-launch: Set start_playwright=True in the client options to automatically launch a Chromium browser that's already configured with the Aluvia proxy. The browser is available via connection.browser and is automatically cleaned up when you call connection.close().

Aluvia API

AluviaApi is a typed wrapper for the Aluvia REST API. Use it to manage connections, query account info, or build custom tooling—without starting a proxy.

AluviaApi is built from modular layers:

┌───────────────────────────────────────────────────────────────┐
│                         AluviaApi                             │
│    Constructor validates api_key, creates namespace objects   │
├───────────────────────────────────────────────────────────────┤
│                                                               │
│   ┌─────────────┐    ┌─────────────┐    ┌─────────────┐      │
│   │   account   │    │    geos     │    │   request   │      │
│   │  namespace  │    │  namespace  │    │  (escape    │      │
│   │             │    │             │    │   hatch)    │      │
│   └─────────────┘    └─────────────┘    └─────────────┘      │
│          │                  │                  │              │
│          ▼                  ▼                  ▼              │
│   ┌────────────────────────────────────────────────────┐     │
│   │              request_and_unwrap / request           │     │
│   │         (envelope unwrapping, error throwing)      │     │
│   └────────────────────────────────────────────────────┘     │
│                            │                                  │
│                            ▼                                  │
│   ┌────────────────────────────────────────────────────┐     │
│   │                   request_core                      │     │
│   │    (URL building, headers, timeout, JSON parsing)   │     │
│   └────────────────────────────────────────────────────┘     │
│                            │                                  │
│                            ▼                                  │
│                      httpx / requests                         │
└───────────────────────────────────────────────────────────────┘

What you can do

Example

from aluvia_sdk import AluviaApi

api = AluviaApi(api_key=os.environ["ALUVIA_API_KEY"])

# Check account balance
account = await api.account.get()
print(f"Balance: {account['balance_gb']} GB")

# Create a connection for a new agent
connection = await api.account.connections.create(
    description="pricing-scraper",
    rules=["competitor-site.com"],
    target_geo="us_ca",
)
print(f"Created: {connection['connection_id']}")

# List available geos
geos = await api.geos.list()
print("Geos:", [g["code"] for g in geos])

Tip: AluviaApi is also available as client.api when using AluviaClient.

License

MIT — see LICENSE