proxyfield
Start with 1 GB
Use case

Claude Proxies: Residential & Mobile IPs

A Claude proxy gives the tools that Claude drives on your side, such as scrapers written in Claude Code, a computer-use browser or an MCP server, residential IPs in the country and city you choose, while your calls to the Claude API keep their direct connection. ProxyField has 412M+ residential and 10.2M+ mobile IPs in 195+ countries and 21,000+ cities, with a 99.6% success rate, according to our service data, from $1.95/GB with no subscription. Start with the Easy Trial: 1 GB for $2.95, no minimum purchase.

Start with the Easy Trial: 1 GB for $2.95Residential proxies
Residential from $1.95/GB · Mobile from $2.45/GB

Targeting covers country, region (state) and city, including all 50 US states. Threads are unlimited, the traffic never expires, and 24/7 live chat helps with setup and integration.

Checkout link: the billing page of your dashboard, proxyfield.io/en/dashboard/billing (sign in first)

gate.proxyfield.io:7000 setup
Type
residential (mobile for the phone view)
Session
a new IP per request for page fetches; sticky from 30 s to 36 h for a browser agent
Geo
the country or city whose pages your agent should see
Residential from
$1.95/GB
Mobile from
$2.45/GB
success rate, according to our service data
99.6%
residential IPs
412M+
mobile IPs
10.2M+
countries
195+
cities
21,000+
  • a new IP per request
  • from $1.95/GB

What is a Claude proxy?

"Claude proxy" names three different things, and only the first is what ProxyField sells:

  • A network proxy for Claude's tools

    The web requests of code and browsers that run on your machine or server go out through another IP. This is a forward proxy, priced per GB.

  • An LLM gateway

    A server, such as LiteLLM or one of the "claude-code-proxy" projects on GitHub, that sits between Claude Code or your app and one or more model APIs. Claude Code reaches one through ANTHROPIC_BASE_URL; it routes, logs or translates the model calls themselves.

  • A company proxy

    The outbound proxy that a corporate network requires. Claude Code reads it from HTTPS_PROXY and HTTP_PROXY.

People who buy residential proxies for Claude are building agents and scrapers: Claude writes or drives the code, and the target sites should see ordinary visitors from the right country instead of one busy server IP.

Where a proxy fits around Claude

Claude's tools run in two places. Those that run on Anthropic's side cannot use your proxy; those that run on your side can.

Claude toolWhere its web requests startCan it use your proxy?
Web search and web fetch (claude.ai and the API)Anthropic's serversNo
Claude Code: scripts it writes and runsYour machineYes, set in the script
Claude Code: its own connection to AnthropicYour machineOnly through HTTPS_PROXY, which you rarely need
Computer useThe computer or container you provideYes, in that browser or system
Claude in ChromeYour ChromeYes, whatever proxy that Chrome profile uses
MCP servers, such as a Playwright browserWhere the server runsYes, in the server's options
Your own tools on the API (tool use)Your codeYes

Open-source agent libraries that drive a browser, such as Browser Use, run on your side too and take the proxy in their browser settings.

The pattern is the same everywhere: give the proxy to the part that fetches pages, not to the model call. Step-by-step scraping with Claude Code, a scraper skill and browser MCP servers: Web Scraping With Claude and Web Crawl and Scraping MCP Servers.

Give a Claude agent's own tool a residential IP

With tool use on the Claude API, the model asks for a tool call and your code runs it. Put the proxy in that code:

Python
# pip install anthropic requests
import os, requests, anthropic

client = anthropic.Anthropic()          # the model call goes straight to Anthropic
PROXY = os.environ["PAGE_PROXY"]        # http://<login>-country-us-city-new_york:<password>@gate.proxyfield.io:7000
MODEL = os.environ["MODEL"]              # the Claude model you use

def fetch_page(url: str) -> str:
    """Runs on your side: only these requests use the proxy."""
    r = requests.get(url, proxies={"http": PROXY, "https": PROXY}, timeout=30)
    return r.text[:20000]

tools = [{"name": "fetch_page", "description": "Fetch a public web page and return its HTML.",
          "input_schema": {"type": "object", "properties": {"url": {"type": "string"}}, "required": ["url"]}}]
messages = [{"role": "user", "content": "Summarize the pricing on https://example.com/pricing"}]

while True:
    msg = client.messages.create(model=MODEL, max_tokens=2048, tools=tools, messages=messages)
    if msg.stop_reason != "tool_use":
        break
    messages.append({"role": "assistant", "content": msg.content})
    messages.append({"role": "user", "content": [
        {"type": "tool_result", "tool_use_id": b.id, "content": fetch_page(b.input["url"])}
        for b in msg.content if b.type == "tool_use"]})
print("".join(b.text for b in msg.content if b.type == "text"))

<login>-country-us-city-new_york in the username asks the gateway for a New York IP. For computer use, set the proxy in the browser inside the computer or container that Claude controls; for a Playwright MCP server, in the server's proxy option. Other agent stacks: AI web scraping.

Which proxy type is best for Claude?

Rotating residential for most agent work: every page fetch gets a new household IP, so a long task does not pile hundreds of requests onto one address. A browser agent that clicks through several pages of one site needs a sticky session, so the site sees one visitor from start to finish.

TypeFit for Claude agentsWhyPrice
Residential, rotatingScrapers and fetch toolsA new household IP per request, 21,000+ cities$2.95 down to $1.95/GB
Residential, stickyBrowser agents and computer useOne IP for 30 s to 36 h$2.95 down to $1.95/GB
Mobile 5G/4GMobile versions of sitesCarrier IPs$3.45 down to $2.45/GB
Datacenter, rotatingSites that accept server trafficLowest cost per GB$0.29 down to $0.22/GB
Dedicated ISP (static residential)A long-lived browser profileThe same IP for 30 daysfrom $2.29 per IP / 30 days

Rotating datacenter proxies cover the same countries as residential; dedicated ISP and datacenter IPs come in the countries their order form lists, which follow stock. Compare: Residential vs Datacenter Proxies, Sticky vs Rotating Proxy Sessions.

What a proxy does not change for Claude

  • API limits

    Anthropic sets rate limits per organization and usage tier. A 429 rate_limit_error or a 529 overloaded_error comes from Anthropic, whatever IP you call from; retry with backoff and honor the retry-after header.

  • Where Claude is offered

    Anthropic publishes the countries where Claude is available, and its terms tie use to them. A proxy changes the IP, not the account's eligibility.

  • Anthropic's own fetching

    Web search and web fetch on claude.ai and in the API leave from Anthropic's network; sites that block Anthropic's bots stay blocked for those tools.

  • Site rules

    A residential IP makes your agent's requests look like a visitor's, not authorized where a site's terms or robots.txt say no. Keep to public pages and a modest pace.

ProxyField has no API that collects Claude's answers. For brand monitoring in AI answers, the AI search scraper API covers ChatGPT, Perplexity and Google AI Mode, from $1.00 per 1,000 successful results at volume (Contact sales). AI search scraper API availability: through Contact sales.

How much do Claude proxies cost?

Per GB: residential $2.95/GB for 1 GB down to $1.95/GB from 1,000 GB, mobile $3.45 down to $2.45. The tier rate applies to the whole purchase, there is no subscription, and the traffic never expires.

Worked example: Claude agents per month (example inputs, not measurements; a page with images and scripts can weigh several MB, raw HTML often a few hundred KB):

SetupVolume a monthTrafficResidential
A test run: a Claude Code scraper on 2,000 pages, 0.5 MB a page2,000 pagesabout 1 GBEasy Trial: 1 GB for $2.95
A daily scraper Claude wrote: 1,500 pages a day, 0.3 MB a page45,000 pagesabout 13.5 GB14 × $2.79 = $39.06
A computer-use agent: 2,000 tasks × 10 pages, 2 MB a page20,000 pagesabout 40 GB40 × $2.79 = $111.60

Blocking images, fonts and video in the agent's browser cuts page weight; the dashboard shows real usage, so measure a sample first. All purchases are final. In return, your traffic never expires. Cheaper than the big brands' $4/GB at every volume: residential proxies cost $2.95/GB for 1 GB and fall to $1.95/GB from 1,000 GB, with no subscription. See Pricing per GB.

How do I set up a proxy for Claude?

Decide which tool's traffic should leave through the proxy, give that tool gate.proxyfield.io:7000 with a username for the country and city, and let Claude fetch an IP-check page through it before the real task.

  1. Get access

    Once your payment is confirmed, the dashboard shows the gateway and your credentials, or lets you allowlist your server's IP.

  2. Make a line per location

    In the configurator, set the country and city, for example <login>-country-us-city-new_york, and a sticky session for browser agents.

  3. Set it on the tool, not on Claude

    In the scraper Claude writes (an environment variable such as PAGE_PROXY), the browser inside the computer-use environment, or the MCP server's proxy option.

  4. Check from inside the agent

    Ask Claude to fetch https://api.ipify.org with the tool, confirm the IP and location, then start the task.

Shell
curl -x gate.proxyfield.io:7000 -U "<login>-country-us-city-new_york:<password>" https://api.ipify.org

HTTP, HTTPS and SOCKS5 work from your first gigabyte. Region (state) and sticky-session syntax: -region-<region>, -session-<id> and -ttl-<seconds> (30 to 129,600 seconds, 20 minutes by default) after the location, as in <login>-country-us-region-new_york-city-new_york-session-he9f2qga-ttl-600. Guides: Python proxy setup, Playwright proxy setup, Chrome proxy setup.

Rules for Claude agents on the web

Anthropic's Usage Policy governs what you build with Claude; ProxyField's acceptable use policy governs the traffic: public web data, monitoring and testing are allowed, while spam, fake accounts, attacks and access to data behind logins you are not entitled to are not. Some site categories, including financial and government services, are restricted on the network. Have your agent read robots.txt, stop on repeated 403 and 429 responses instead of retrying in a loop, and collect only the fields you need. More: Is AI Scraping Legal?

FAQ

Claude proxies FAQ

Can't find your answer? Support is on 24/7 live chat, including help with setup and integration.

What is a Claude proxy?

ProxyField's Claude proxies are residential and mobile IPs, 412M+ residential in 195+ countries and 21,000+ cities from $1.95/GB, for the scrapers, browsers and tools that Claude drives on your side. The name also covers LLM gateways and company proxies, which handle the model calls rather than the page fetches.

What is a Claude API proxy?

Usually one of two things: a network proxy that your app's requests leave through, or a gateway server that receives Claude API calls and forwards them, for logging, key management or routing between models. ProxyField is the first kind and is meant for the web requests your agent makes, not for the API call.

Why use a proxy for Claude?

So that the pages a Claude agent or a Claude-written scraper fetches come from residential IPs in the right country, spread over many addresses instead of one server. It also lets you see sites as visitors in another city see them.

Which proxy is best for the Claude API?

ProxyField residential proxies cost $2.95/GB at 1 GB, down to $1.95/GB from 1,000 GB, and are meant for the tools around the Claude API, not for the API call. The call itself needs no special proxy: Anthropic identifies you by your API key. For the tools, rotating residential IPs suit page fetches and sticky residential sessions suit browser agents.

How do I set up a Claude API proxy?

To route the client library's own calls, which only a network that requires it needs, set HTTPS_PROXY or give the Anthropic client an HTTP client with a proxy. For an agent's page requests, put the proxy in the tool function instead, as in the example on this page.

Can I put Claude Code behind a proxy?

Yes. Claude Code reads the standard HTTPS_PROXY, HTTP_PROXY and NO_PROXY variables for its connection to Anthropic, which is what company networks need. For scraping, leave Claude Code's own connection alone and give the proxy to the scripts it runs.

Should I use sticky sessions or per-request rotation for Claude API use?

For the API call it makes no difference. For an agent's tools, rotate per request when each call fetches a different page, and hold a sticky session, from 30 s to 36 h, when a browser agent works through several pages of one site.

Are free proxies good for the Claude API?

No. A free proxy sits between your app and Anthropic and can log the traffic, may be slow or offline, and is often blocked by target sites. Use your direct connection for the API and a paid proxy for page fetches.

Can a proxy fix Anthropic API error 429?

No. A 429 rate_limit_error means your organization went over its rate limit for that tier, and a new IP leaves the limit where it is. Retry after the time in the retry-after header, spread requests out, or move to a higher usage tier.

Can a Claude API proxy fix 403 Forbidden errors?

Not from the Claude API: a 403 there is a permission error tied to your key, organization or account. A 403 from a website your agent visits is the site refusing the request; slow down, and check the site's terms and robots.txt before you retry. More: 403 Forbidden.

Show 13 more questionsShow fewer questions

Can I use a Claude API proxy for Anthropic API key rotation?

ProxyField does not handle API keys; it only changes the IP your requests come from. Rate limits are set per organization, so rotating IPs, or keys within one organization, does not raise them.

Do I need a Claude reverse proxy or a forward proxy?

A forward proxy, such as ProxyField, changes the IP that your requests leave from. A reverse proxy or gateway is a server in front of the API that manages keys, logs calls or switches models; it sees every prompt, so run your own rather than a public one.

Can I use Claude Code with other models through a proxy?

That is an LLM gateway's job, not a network proxy's: point ANTHROPIC_BASE_URL at a gateway that speaks Anthropic's API format and forwards the calls to the models it supports. A residential proxy only changes the IP your requests come from.

Can I use Claude through a proxy from a country where it is not available?

Anthropic publishes the countries where Claude is available, and a proxy changes your IP, not your account's eligibility. A proxy around Claude is for the web requests of your agents and scrapers, so the sites they fetch see a visitor from the country you choose.

Can Claude's browser use a SOCKS5 residential proxy?

A browser you run, such as Chromium or Firefox driven by Playwright, accepts SOCKS5 as well as HTTP proxies. Chromium does not accept a username and password for SOCKS5, so the HTTP proxy at gate.proxyfield.io:7000 with your credentials is the setup to use there; SOCKS5 runs on the same gate.proxyfield.io:7000 with the same username and password.

How do I verify that Claude's browser used the proxy?

Have Claude open https://api.ipify.org or another IP-check page in that browser and read the result back. It should show a proxy IP in the location you set, not your server's IP.

Can multiple Claude agents share one browser context?

They can, but they then share cookies, sessions and one IP, so one agent's actions affect the others. Give each agent its own browser context with its own sticky session, which also keeps its traffic separate in your logs.

Is Claude Browser Use the same as the Browser Use Python package?

No. Browser Use is an open-source Python library that lets several models, Claude among them, drive a Playwright browser; Anthropic documents its own browser and computer-use tools separately. In both cases the browser runs on your side and takes your proxy.

How do I give Claude live web access?

Turn on web search or web fetch, which run on Anthropic's side, or connect tools that run on yours: a fetch function, a browser through computer use or an MCP server. Only the tools on your side can use a residential proxy in the country you choose.

How much does it cost to add web access to Claude?

Anthropic charges for its web search tool per search on top of tokens. With your own fetch tool or browser, the web traffic costs what your proxy costs: ProxyField residential from $1.95/GB, with the Easy Trial at 1 GB for $2.95.

Why does Claude use old data instead of current prices?

Without a web tool, Claude answers from its training data, which ends at a cutoff date. Give it web search or a fetch tool, and through a proxy in the right country it reads the prices shown there today.

What data can I extract from Claude Web?

ProxyField has no API that collects Claude's answers. Its AI answer APIs cover ChatGPT, Perplexity and Google AI Mode, at $1.50 per 1,000 successful results, down to $1.00 at volume; see the AI search scraper API.

How much do Claude proxies cost?

Residential traffic costs $2.95/GB for 1 GB down to $1.95/GB from 1,000 GB, with no subscription. A Claude-written scraper fetching 1,500 pages a day at 0.3 MB a page uses about 13.5 GB a month, $39.06 for 14 GB.

Keep reading

Related pages

Easy Trial: 1 GB for $2.95.

Try for the price of 1 GB, no subscription, traffic never expires.

Start with 1 GB See pricing