Loading page content.
Dedicated IPv4 out of our own ARIN and RIPE allocations, unmetered on 10 Gbps ports. The lowest cost per successful request on any target that is not scoring ASNs.
Starting price
$2.00/IP/wk
per IP, week, from 1 IP. Falls to $1.10/IP/wk at 500 IPs.
Every account starts with 50 MB of free residential bandwidth. No card is required to test the network.
How it works
These are not resold subnets. FleetProxy holds its own ARIN and RIPE allocations and announces them from fourteen regions, which means the routing, the reputation and the abuse handling are ours to fix rather than a supplier's to escalate.
Addresses come from FleetProxy's own ARIN and RIPE allocations in 14 regions, so nothing is resold and route problems are ours to fix rather than a supplier's. Each IP is held by one customer for the length of the lease and moves unmetered traffic on a 10 Gbps port. Latency to the major clouds in the same region stays in single-digit milliseconds.
You choose a region at checkout and an address is allocated from our range there. It is dedicated for the term — nobody else is placed in your /24 and no neighbour's traffic touches your reputation.
There is no gateway and no pool matching. Your client opens connections straight to the leased address, so the only latency between you and the target is the network path.
Each address sits on a 10 Gbps port with unmetered transfer. Concurrency is limited by your own client and by what the target tolerates, not by a thread quota or a bandwidth counter.
Latency to AWS, GCP and Azure in the same region stays in single-digit milliseconds. For high-volume crawling and load testing that adjacency matters more than any other property of the address.
The honest limitation: this is a hosting address and it is identifiable as one. Any target buying commercial IP-intelligence data can see a datacenter ASN in one lookup. Datacenter proxies are the right tool where nobody is looking, and the wrong tool the moment somebody is.
Volume ladder
Tiers are a step function, not interpolation: the greatest break at or below your quantity wins and every unit bills at that rate. They apply automatically at checkout — there is no code to enter and nothing to negotiate.
| Quantity | Unit price | Saving | Order total | Order at this tier |
|---|---|---|---|---|
| 1 IP and above | $2.00/IP/wk | Entry rate | $2.00 | Order |
| 5 IPs and above | $1.90/IP/wk | Save 5% | $9.50 | Order |
| 25 IPs and aboveMost popular | $1.80/IP/wk | Save 10% | $45.00 | Order |
| 50 IPs and above | $1.70/IP/wk | Save 15% | $85.00 | Order |
| 100 IPs and above | $1.50/IP/wk | Save 25% | $150.00 | Order |
| 250 IPs and above | $1.30/IP/wk | Save 35% | $325.00 | Order |
| 500 IPs and aboveBest value | $1.10/IP/wk | Save 45% | $550.00 | Order |
Totals are what checkout charges for exactly that quantity. Buying past a break lowers the rate on every unit, not only the ones above it. Price a custom quantity in the calculator.
Specification
Everything a reader needs to decide whether this product fits their stack, on one screen and in one table.
| Property | Value |
|---|---|
| Rotation | None. The address is dedicated for the lease; replacements are one click |
| Sticky duration | The whole term, across restarts and redeploys |
| Targeting | Region — 14 of them, including Ashburn, Frankfurt, Singapore and São Paulo. No city or ASN selection: the ASN is ours |
| Protocols | HTTP and HTTPS via CONNECT on port 8080 and 8443, SOCKS5 on 1080, direct to the address |
| Concurrency | Unlimited connections on a 10 Gbps port. No per-thread charge and no connection cap |
| Bandwidth policy | Unmetered transfer for the length of the lease |
| Authentication | Username and password, or an IP whitelist, over any supported protocol |
| Order size | 1 IP to 2,000 IPs per order, in steps of 1 |
Integration
The address is the endpoint and the region was fixed when it was allocated, so the username carries no segments. This is the simplest credential we issue and the one that drops most cleanly into existing infrastructure.
fp_a41c9dfp_a41c9d
The credential id for this address.
198.51.100.7
The leased address, announced from the region you chose. It is what the target logs.
IP whitelist
Add your egress addresses to the credential and the proxy accepts connections from them. The password is still required.
No -country segment
Region was chosen at allocation. Buying a second region means a second address.
# 200 workers against one dedicated address. There is no thread ceiling
# and no per-connection charge; the 10 Gbps port is the only limit.
xargs -a urls.txt -P 200 -I{} \
curl -s -o /dev/null -w '%{http_code} %{time_total}\n' \
-x http://fp_a41c9d:[email protected]:8080 {}Targeting
One dimension: where the address is announced from. There is no city or ASN selection, because every address in this product sits on the same autonomous system — ours.
| Control | How it is set | Notes |
|---|---|---|
| Region | Checkout selection | Fourteen regions across North America, Europe, Asia-Pacific and South America. |
| City | Region-level only | A region maps to a specific facility; the city list below shows where the racks are. |
| ASN | Fixed | Traffic leaves on our own allocation, which is what makes route problems ours to fix. |
| Replacement | One click | Swap an address that has picked up a block; the replacement is announced from the same region. |
Datacenter regions with published inventory. Racks, not peers — the counts below are addresses we announce ourselves.
See every country and city with live inventory — counts are the addresses published for that market, not a global total repeated on every page.
Use cases
Four workloads where this product is the right answer rather than a workable one.
Open data, documentation, sitemaps, government registers and anything else that does not defend itself. At datacenter rates the cost per million pages is a rounding error next to residential, and unmetered transfer removes the bandwidth budget from the design entirely.
Driving synthetic traffic at your own infrastructure from a fixed, known address you can allowlist at the far end. Single-digit millisecond latency to the major clouds in-region keeps the measurement about your service rather than about the path.
Crawling client sites, auditing redirects, checking canonical tags and rendering budgets. The target is usually your own customer, so ASN reputation is irrelevant and throughput is everything.
Latency-sensitive workloads where a stable, dedicated, low-latency address matters more than looking like a consumer. SOCKS5 on the same lease covers protocols that are not HTTP.
Browse every use case for the full breakdown, including the products that suit each one.
Compare
The two nearest products in the catalogue, and the conditions under which each of them is the better purchase. Buying the wrong class costs more than buying the wrong volume.
$1.20/GB
Choose rotating datacenter instead if you would rather have breadth than a fixed address: 240,000 exits behind one endpoint, billed per gigabyte, with no lease to size in advance. A crawl that trips per-IP rate limits wants many addresses, not one fast one.
$3.50/IP/wk
Choose static residential instead the moment the target scores ASN reputation. No amount of bandwidth fixes a hosting address on a site that filters hosting addresses — the ISP lease costs more per week and is the only one of the two that gets a response.
Compare all eight proxy types on identity, rotation, targeting and billing, or price them side by side.
Questions
Pulled from the help center and filtered to the questions this product actually raises.
You can add an IP whitelist to any credential, but password authentication is still required alongside it — the whitelist narrows who may use the credential, it does not replace authentication. Whitelists are per-credential rather than per-account, so adding an address to one credential does not add it to the others. A changed office IP or a new CI runner is a common cause of sudden 407s.
A 407 comes from our gateway, not the target, and the x-fleet-reason header on the response says which of the six causes it is: bad-credentials, no-matching-pool, ip-not-whitelisted, quota-exhausted, or plan-inactive. The most common causes are a client that drops credentials from the proxy URL and a password containing a character that needs percent-encoding. Reproduce with curl first; if curl works, the fault is client-side.
Residential IPs are addresses assigned by consumer ISPs to real subscriber connections, so targets see an ordinary visitor; they cost more and add 200–900ms of latency. Datacenter IPs come from hosting providers, are roughly twenty times cheaper per gigabyte and much faster, but are trivially identifiable and filtered by any site that cares. Mobile IPs are carrier-assigned 4G/5G addresses shared by thousands of subscribers, which makes them the hardest to block and the most expensive. Start with datacenter on undefended targets and residential everywhere else.
Yes, on port 1080 with username and password authentication. Use the socks5h scheme rather than socks5 so DNS resolves at the proxy — otherwise your local resolver sees every hostname and geo-aware CDNs resolve against your real location, which defeats country targeting. Port 8080 handles HTTP and HTTPS via CONNECT.
Roughly 30 million residential addresses across 195 countries, plus dedicated static residential, mobile and datacenter inventory. Coverage is not uniform: the locations pages show live IP counts per country and city so you can check a specific market before ordering rather than after. Anything under 500 IPs in a pool is flagged, because a pool that small will repeat addresses inside a large crawl.
Something not covered here? Read the full help center or ask the support team.
Start on 50 MB of free residential bandwidth, then buy from $2.00/IP/wk. Volume tiers apply automatically, and there is no contract to sign.
No card required for the trial. Cancel or downgrade at any time.