Agent37是Y Combinator F26批次的初创公司，致力于为运行AI代理提供成本极低的沙箱基础设施。1该公司的核心竞争力在于将服务成本压低至行业标准的不足10%，用户平均月费仅为$0.37/沙箱。1免费沙箱的基础设施成本平均约为$0.05-$0.10/用户/月。1
公司创始人Vishnu曾担任Amazon高级软件工程师和Quantcast工程经理，主要参与构建Amazon内部AI聊天机器人Diya和编码工具Ziya。1Agent37通过多项技术手段实现了显著的成本优化：采用gVisor技术实现跨沙箱资源缓存和共享以支持快速启动，高效管理AI代理的突发CPU和RAM使用模式，选择Hetzner和OVH等替代云服务商而非超大规模云商，以及通过优化服务器配置侧重更大RAM容量。1在用户端，新注册用户中超过90%采用免费计划。1
Agent37, a Y Combinator F26 startup, is introducing agent sandboxes priced at a fraction of industry standard costs 1. The platform charges users an average of $0.37 per sandbox monthly, representing less than 10% of typical market rates 1. The company achieves this dramatic cost reduction through a combination of technical optimizations and infrastructure choices, including efficient CPU and RAM management for AI agent workloads, deployment on alternative cloud providers like Hetzner and OVH rather than hyperscale vendors, implementation of gVisor technology for cross-sandbox resource caching and rapid startup, and server configurations prioritizing larger RAM capacity 1.
The service has already attracted significant user interest, with free-tier users comprising over 90% of new registrations 1. For free users, the underlying infrastructure cost averages between $0.05 and $0.10 per user monthly 1. Founder Vishnu brings relevant experience from his tenure as a senior software engineer at Amazon, where he contributed to building the company's internal AI chatbot Diya and coding tool Ziya, and from his prior role as an engineering manager at Quantcast 1.
评论
还没有评论，欢迎留下第一条。