ISP Proxies in United States
Datacenter-hosted IPs announced under real residential ASNs. Stable, persistent, and the right tool when cookies and session state need to survive across a multi-turn evaluation or a sustained RAG ingestion.
- IP allocation
- Static, dedicated
- ASN class
- Residential-announced
- Latency
- < 50ms regional
- Bandwidth
- Unmetered
- Protocols
- HTTP, HTTPS, SOCKS5
- Rotation
- None (explicit reset only)
- Uptime SLA
- 99.95%
United States · ISP
US ISP proxies for long-session AI workloads and authenticated RAG
Where US ISP is the right choice
ISP proxies sit between residential (high authenticity, high rotation overhead) and datacenter (high throughput, low authenticity) on the AI workload tradeoff curve. The US ISP pool is the right tool for four specific AI workloads:
- Multi-turn agent evaluation on US-region model APIs. Agent benchmarks that run 20-50 turns per session need stable IP for the life of the conversation. Residential rotation breaks session cookies; ISP static keeps the session coherent.
- Authenticated RAG ingestion from US enterprise wikis. Notion public exports, Confluence public, and similar session-cookied sources need persistent IP through the browse traversal.
- Sustained HuggingFace dataset pulls. HF's rate limiter is per-IP; a stable ISP session with sticky behaviour lets LFS resumption work reliably across TB-scale dataset pulls.
- Long-session scraping of US academic platforms. OpenReview, some ACM / IEEE portals that require a trust-tier that accepts residential-announced ASN but wants consistent IP through the session.
What the US ISP pool gives you
- Static allocation — your IP stays assigned for the life of the subscription unless you explicitly request rotation.
- Residential-ASN announcement — IPs announce under real residential ASNs (Comcast, Spectrum, Verizon) from AWS-hosted infrastructure. Targets see a residential origin; bandwidth performs like datacenter.
- Sub-50ms regional latency — hosted in AWS us-east-1 and us-west-2; latency-sensitive eval workloads anchor here.
For the country context, see the US country page; for the ISP class across all 10 countries see the ISP proxy page.
Why ISP exists between residential and datacenter
ISP proxies are datacenter-hosted IPs that have been announced under a real residential ASN. You get the persistence and latency of datacenter with the ASN reputation of residential. The trade-off: no diversity — you have a specific, stable set of IPs assigned to your account.
For AI workloads, ISP is specifically useful when:
- A multi-turn agent evaluation depends on the same IP across a long conversation, because mid-session IP rotation breaks session cookies, CSRF tokens, and some chat-history features on commercial APIs
- A RAG ingestion pipeline logs into a knowledge source (with permission) and the source's rate limiter is per-IP
- You need sub-50ms latency to a US or EU eval target but the target ASN-classifies AWS subnets and downgrades you
Where ISP is wrong
- SquadProxy diversity. A training-corpus workload sampling broadly from the open web does not benefit from ISP's persistence. Use residential or datacenter instead.
- Mass parallel evaluation. Running 50k concurrent eval requests from 100 ISP IPs means 500 requests per IP — which is not residential-shaped behaviour even though the IP has a residential ASN. Mix with rotating residential above scale.
Regions and ASN mapping
- US — Comcast, Spectrum, AT&T, Verizon.
- UK — BT, Sky, Virgin Media, TalkTalk.
- Germany — Deutsche Telekom, Vodafone DE, 1&1.
- France — Orange, Free, SFR, Bouygues.
- Japan — NTT, KDDI, SoftBank.
- Singapore — Singtel, StarHub, M1.
Korea, Canada, Netherlands, and Australia ISP allocations available on request but are smaller pools. Ask before quoting.
Provenance
ISP announcements are done through upstream partnerships with carriers that lease residential IP space. Every allocation has a RIR-visible ownership chain ending at the carrier. For customers publishing model cards or data cards that have to disclose source infrastructure, this is documentable.
Pricing
Pricing for United States isp
Every plan includes the United States isp pool — Static, dedicated of coverage, with sticky sessions up to 60 minutes.
Solo
For individual researchers running evaluation scripts and prototype RAG pipelines.
$149/ month
or $1,430/year (save 20%)
50 GB residential · unlimited datacenter · 200 concurrent sessions
- ✓Access to all 5 exit classes · 10 focus countries
- ✓50 GB residential · unlimited datacenter
- ✓5 static ISP IPs · 5 GB 4G mobile
- ✓1 seat · 200 concurrent sessions
- ✓Python + Node SDK + REST API
- ✓Per-request metering (not time-based)
- ✓Email support (24h response, business days)
- ✓Overage: $3/GB residential · $6/GB mobile
Best for
- Solo researchers
- Evaluation scripts
- Prototype RAG
Team
Most popularFor AI startups and mid-size labs splitting capacity between training and evaluation.
$699/ month
or $6,710/year (save 20%)
500 GB residential · unlimited datacenter · 1,000 concurrent sessions
- ✓Access to all 5 exit classes · 10 focus countries
- ✓500 GB residential · unlimited datacenter
- ✓25 static ISP IPs · 25 GB 4G mobile
- ✓10 seats ($29/mo per extra seat) · 1,000 concurrent sessions
- ✓City-level geo-routing + ASN targeting
- ✓99.9% uptime SLA
- ✓Priority Slack support (4h response, business hours)
- ✓Python + Node SDK + REST API + webhooks
- ✓Overage: $3/GB residential · $6/GB mobile
Best for
- AI startups
- Mid-size labs
- Model eval teams
Lab
For academic labs, eval consortia, and frontier model companies running sustained workloads.
$2,999/ month
or $28,790/year (save 20%)
2 TB residential · unlimited DC · 50 GB 4G + 20 GB 5G · 3,000 concurrent sessions
- ✓Access to all 5 exit classes · 10 countries on 4 continents
- ✓2 TB residential · unlimited datacenter
- ✓100 static ISP IPs · 50 GB 4G + 20 GB 5G mobile
- ✓50 seats ($19/mo per extra seat) · 3,000 concurrent sessions
- ✓Dedicated gateway lane (bypasses shared-pool queues on us-east-1 + eu-west-1)
- ✓99.95% uptime SLA
- ✓Dedicated Slack channel (1h response, business hours)
- ✓Custom BGP prefix on request (additional fees apply)
- ✓Overage: $2.50/GB residential · $5/GB mobile
Best for
- Academic labs
- Large eval consortia
- Frontier model companies
Enterprise
Custom contracts with dedicated infrastructure, volume pricing, and research-grade SLAs.
Custom pricing
Custom (from 5 TB/mo residential) · unlimited concurrent sessions
- ✓Volume pricing from 5 TB/mo residential
- ✓Dedicated BGP prefix + ASN announcement
- ✓Unlimited concurrent sessions · unlimited seats
- ✓99.99% uptime SLA with financial credits
- ✓Named Technical Account Manager + 24/7 on-call paging
- ✓Custom AUP, DPA, on-site deployment option
- ✓Research / academic discount (30–50% off Team or Lab)
- ✓Annual contract · wire, ACH, USDC/USDT/BTC settlement
Best for
- Frontier labs
- Eval consortia
- Enterprise AI
All plans include 14-day refund, single endpoint with regional failover, HTTP(S) + SOCKS5 on every exit class, access to all 5 exit classes and all 10 focus countries, and Python + Node SDKs. Concurrent sessions = simultaneous TCP sessions through the gateway. Overage warnings fire at 80% and 100%; traffic continues only if overage billing is enabled on your account.
FAQ
United States ISP FAQ
Are these really ISP proxies in United States?
Yes — the exits in this pool announce from the United States ASNs we list on the country page (Comcast (AS7922), Charter Spectrum (AS20115)) and route through our North America edge. You can verify ASN and carrier on every request.What session lengths can I hold on United States ISP?
Per-request rotation is the default. Sticky sessions are supported up to 60 minutes via the X-Squad-Session header. For anything longer, use our ISP pool instead.Is targeting down to a city available on United States ISP?
Yes, via X-Squad-City. Pool depth varies by city — the deepest are New York, San Francisco, Seattle. Smaller cities are served but have thinner concurrency.Any legal or compliance notes specific to United States?
Scraping through US residential and datacenter exits sits within a settled legal landscape around CFAA (post-hiQ v. LinkedIn), copyright (Authors Guild v. Google), and state-level consumer privacy regimes (CCPA/CPRA, Virginia, Colorado, Texas). AI training-data use cases — particularly those relying on "publicly available" content — should be reviewed under the Copyright Office's evolving guidance on generative AI. Our AUP prohibits circumvention of technical access controls and scraping behind authentication.
Start routing United States traffic through isp
Real ASNs, real edge capacity, and an engineer who answers your Slack the first time.