A new proxy service launched this week with a simple pitch: AI agents need internet access, and the existing infrastructure for providing it is built for humans. x402socks.com sells SOCKS5 proxy access that requires no accounts, no email, and no identity verification. Payment happens entirely through cryptocurrency.
The service is part of a growing category of infrastructure designed specifically for autonomous software agents, the kind that OpenAI, Anthropic, and other companies are increasingly deploying to perform multi-step tasks on the internet. These agents can browse websites, query APIs, and interact with web services, but they face a practical problem: most proxy services require human-style account creation, email verification, and payment methods that tie back to a person.
The Agent Access Problem
AI agents that need to access external resources run into friction at every step. A proxy service typically asks for an email address, credit card information, and sometimes phone verification. An agent handling a research task or data collection workflow doesn't have any of these. It can create a temporary email, but that introduces delays, and many services block disposable addresses anyway.
The x402 Protocol, which x402socks.com uses for payments, was built to solve this. It's a machine-to-machine payment standard that lets software pay for services directly using stablecoins. An agent can hold a wallet, receive funds, and pay for proxy access without ever creating an account or revealing identity information. The name comes from the HTTP 402 status code, which was reserved for "Payment Required" but never widely implemented until now.
The service accepts USDT and USDC on Ethereum, Base, and BNB Chain. An agent pays, receives a token, and connects to a SOCKS5 proxy endpoint. No forms, no verification, no waiting.
Built for Machine Consumption
The technical design reflects its target audience. x402socks.com publishes an llms.txt file, a standard that lets AI agents discover available endpoints and documentation without scraping HTML pages. The API follows conventional REST patterns, and the proxy itself uses standard SOCKS5 with minimal protocol overhead.
Regional exit nodes are available, and the service emphasizes low latency. For agents performing tasks like web scraping, API testing, or data retrieval, connection speed and geographic distribution matter more than they do for a human browsing the web casually.
The company frames the service as a tool for developers building autonomous agent systems. The example documentation shows a curl command that an agent could execute to route traffic through a US East exit node, using a token purchased with cryptocurrency.
What the Pricing Looks Like
The service offers both time-based and usage-based plans. The entry-level option costs $1 for six hours of access with a 1 GB data cap. A full day costs $2.49, and a week runs $9.99. For agents with predictable data needs, usage-based plans are available: 5 GB for $6.99 or 50 GB for $34.99, with no time limit on either.
A higher-tier plan at $29.99 provides 30 days of shared access for up to five agents, suggesting the service anticipates teams running multiple concurrent workloads.
All plans include a fair use policy. The service defines this as roughly 95% of 1 Mbps sustained throughput, which translates to about 1 GB per hour, 10 GB per day, 75 GB per week, or 300 GB per month. The language is unusual for a proxy service but makes sense in context: agents can sustain connections for long periods without human attention, and the fair use limits set expectations for what "unlimited" means in practice.
Why This Matters for AI Infrastructure
Proxy services are mundane infrastructure, but the specific combination of features here points to a broader shift. The traditional model of internet access assumes a human user at the other end: someone who can read terms of service, enter a credit card, solve a CAPTCHA, and manage account settings. Autonomous agents can't do any of this reliably, and the friction adds up quickly when an agent needs to make hundreds of requests across different services.
The x402 Protocol addresses a real gap. Stablecoins on public blockchains provide a payment rail that any software can access without permission. An agent with a funded wallet can pay for services the same way a human swipes a card, but without the identity layer that makes card payments possible. This creates a parallel economy where machines transact directly with machines.
The privacy angle is also significant. No logs means the service operator can't reconstruct what an agent did while connected. For developers building agents that handle sensitive data, this is a meaningful feature. It also raises questions about accountability if an agent uses the proxy to do something harmful, a tension that runs through much of AI infrastructure development.
The Bigger Picture
x402socks.com is a small service, but it sits at the intersection of several larger trends. AI agents are becoming more capable and more autonomous. The infrastructure layer that supports them is starting toεε from human-oriented services. And cryptocurrency, long criticized for lacking practical utility, is finding a genuine use case in machine-to-machine payments where traditional banking systems can't operate.
The service also reflects a reality that most AI safety discussions overlook: agents already need practical tools to function in the world. They need network access, payment methods, and ways to interact with web services. Building these tools with privacy and anonymity by default means agents can operate without leaving identity traces, which is useful for legitimate automation tasks and potentially dangerous for everything else.
For developers working on agent systems, x402socks.com represents a straightforward utility. For the broader AI ecosystem, it's an early example of what infrastructure designed entirely for machines, rather than humans, actually looks like.