Skip to content

AI / ML teams

Proxies for agents, crawlers and training data

The same six pools the rest of the site sells, pointed at the jobs AI teams actually run. Plain HTTP and SOCKS5, so it drops into whatever framework you are already using.

6
Proxy pools to route agents through
3
Protocols: HTTP, HTTPS, SOCKS5
Per GB
Billing, with no per-request fees
0
SDKs you have to install

Workflows

Six jobs teams point us at

None of these need a different account or a different gateway. They are the same credentials with a different pool behind them.

Autonomous agent browsing

Give every agent run its own sticky exit so one blocked session does not take the rest of the fleet down with it. Session IDs go in the username, so you can mint a new identity per task without touching your account settings.

Training corpora

Pull large text and image sets without hammering one address. Prepaid residential traffic lasts twelve months, so a collection run that pauses for a month picks up where it left off at the same rate.

RAG refresh pipelines

Re-crawl the sources behind your index on a schedule, from the country your users are actually in. Stale embeddings usually come from a crawl that quietly started failing, not from the model.

Geo-aware evals

Run the same benchmark from twenty markets and compare what the model is really being served. City and ASN targeting on Premium means the eval sees what a local user sees.

Headless browser fleets

Playwright and Puppeteer both take a proxy per browser context, so a hundred workers can hold a hundred separate identities on one balance.

Getting past anti-bot

When a target starts handing back CAPTCHAs, move that worker onto the premium pool and keep going. Same credentials, same gateway, different class of IP.

Drop-in

It is just a proxy URL

Nothing to install and no client of ours to keep updated. If it takes a proxy, it takes this.

python
# Any HTTP client works. No wrapper, no SDK.
# City targeting runs on Premium or Premier residential.
import httpx

PROXY = (
    "http://user-country-us-city-dallas-ttl-10-session-agent01:"
    "YOUR_PASSWORD@gw.foidpro.xyz:7000"
)

with httpx.Client(proxy=PROXY, timeout=30) as client:
    r = client.get("https://example.com/pricing")
    print(r.status_code, len(r.text))
playwright
// One identity per browser context.
const browser = await chromium.launch({
  proxy: {
    server: "http://gw.foidpro.xyz:7000",
    username: "user-country-de-session-eval7",
    password: "YOUR_PASSWORD",
  },
});

const page = await browser.newPage();
await page.goto("https://example.com");

Agents

  • OpenAI Agents SDK
  • Anthropic Computer Use
  • LangChain
  • LlamaIndex
  • CrewAI

Browsers

  • Playwright
  • Puppeteer
  • Selenium
  • Patchright

Crawlers

  • Scrapy
  • Crawlee
  • Apify
  • Octoparse

Clients and glue

  • httpx
  • requests
  • axios
  • cURL
  • n8n
  • Make

What the proxy layer has to do

Four things that matter once you are running at scale

Everything else is marketing. These are the parts that decide whether a fleet of workers keeps running overnight.

Standard protocols, nothing proprietary

HTTP, HTTPS and SOCKS5. If your framework accepts a proxy URL, it already works. There is no wrapper to install and no version of ours to keep up with.

Session control per worker

Rotate on every request, or hold an exit for up to two hours on residential, up to a day on a plan, and for the whole session on ISP. The session ID is a string you pick, up to 32 letters and digits.

Targeting down to the ZIP

Country on every pool, state, city and ASN on Premium and Premier, and ZIP on Premium, set per request. Useful when an eval needs to run as a user in Dallas rather than a user somewhere in the United States.

Pricing you can model

Per gigabyte on every pool, with the rate falling as volume grows. No per-request fees, and any overage on Pro and Scale is billed at your plan rate.

Getting started

From nothing to a first request

Four steps, and the only one that takes any thinking is picking the pool.

  1. 01

    Create an account

    Email and a password, for anyone 18 or over. Ask on Discord for 250 MB of premium residential to test with before you spend anything.

  2. 02

    Pick a pool and a region

    Standard for bulk crawling, premium when the target fights back, ISP when an agent needs to keep one identity for weeks.

  3. 03

    Paste the proxy URL

    Drop it into your client, your browser launch options or your framework config. Targeting and session IDs live in the username string.

  4. 04

    Watch it in the dashboard

    Usage per sub-user, so you can see which agent is burning traffic before the invoice tells you.

Pools

Which one to reach for

Most AI work starts on premium residential and only moves once a specific target pushes back, or once the bill says it should move down.

Rate
from $3.50 / GB
At scale
$2.50 / GB
You pay for
Bandwidth you use
Pool
Rotating residential
Rotation
Per request, or sticky up to 2 h (up to 24 h on a plan)
Protocols
HTTP, HTTPS, SOCKS5
Targeting
Country
Typically used for
Bulk crawlingMarket researchPrice monitoringDataset building
Rate
from $5.50 / GB
At scale
$3.93 / GB
You pay for
Bandwidth you use
Pool
Rotating, re-scored daily
Rotation
Per request, or sticky up to 2 h (up to 24 h on a plan)
Protocols
HTTP, HTTPS, SOCKS5
Targeting
CountryStateCityZIP codeASNSpeedQualitySatelliteIPv6OS
Typically used for
Web scrapingSEO monitoringAd verificationRetail & ticketing
Rate
from $7.00 / GB
At scale
$5.00 / GB
You pay for
Bandwidth you use
Pool
Carrier-issued mobile
Rotation
Per request, or sticky up to 2 h (up to 24 h on a plan)
Protocols
HTTP, HTTPS, SOCKS5
Targeting
Country
Typically used for
Social mediaApp testingAccount managementAd verification
Rate
from $4.00 / GB
At scale
$2.86 / GB
You pay for
Bandwidth you use
Pool
Datacenter
Rotation
Static or rotating subnets
Protocols
HTTP, HTTPS, SOCKS5
Targeting
CountryCitySubnet diversity
Typically used for
High-volume crawlingUptime monitoringAPI accessInternal tooling

Every pool runs the same 1 to 500 GB ladder, so the rate falls at the same points whichever you pick. Published rates run down to the biggest pack size. If you are moving terabytes a month, talk to us and we will quote the run by hand. A monthly plan is usually cheaper again.

Ask for a bulk quote

Plans

Corpus runs belong on a plan

Collection work is steady and large, which is exactly where a monthly allowance beats paying by the gigabyte. Scale and Enterprise add more overage headroom, invoice billing and a named account manager.

Questions

What AI teams ask us first

If yours is not here, ask on Discord or open a ticket and a person answers.

No. Everything speaks plain HTTP, HTTPS and SOCKS5. Any framework that takes a proxy URL will work, including ones that came out after you read this.

Yes. Add a session ID to the username and that worker keeps the same exit until the sticky window ends or you change the ID. There is no cap on concurrent sessions, so one identity per worker scales with the fleet.

Retry on a new session ID and you land on a different exit. If a whole target starts refusing you, switch that worker to Premier, which scores every session, or to Premium with quality-high in the username.

Yes. Lock the country in the username on any pool, or the state, city or ASN on Premium and Premier, and every request from that worker exits there, so results stay comparable across runs.

We count bandwidth and record authentication events for billing and abuse response. We do not log destination URLs or payloads. The Privacy Policy sets out exactly what is kept and for how long.

Start on premium residential. Move the unprotected endpoints down to standard or datacenter to cut the bill, and keep premier for the handful of sources that block everything else.

Test it against your own targets

Ask on Discord or in a ticket for 250 MB of premium residential. No card, and nobody will try to book a call with you.

Accounts are for adults 18 and over. Buying means agreeing to the Terms and Acceptable Use Policy.