Maritime is building a managed cloud for production AI agents, arguing that isolated compute sandboxes address only a small part of the infrastructure teams need. After speaking with hundreds of companies developing agents - from personal assistants and coding tools to healthcare products - the founders found that customers still had to build substantial systems around sandbox providers. Those systems include starting and stopping virtual machines, preserving state, scheduling work, routing users, setting spending limits, monitoring health, recovering from failures, and updating fleets without downtime. Agents also require diverse environments, such as Windows or Android, persistent browser sessions, custom software and operating systems, private-network access, residential IPs, or tightly restricted networking.
Maritime combines each agent’s isolated, persistent microVM with configurable environments and services for provisioning, storage, sleep and wake orchestration, triggers, scheduling, monitoring, networking, updates, rollbacks, recovery, spending tracking, and login handoffs. It also emphasizes lowering costs through scale-to-zero, shared base images, compressed memory snapshots, and tiered storage. The company says a 2-GB agent costs $1 per month for unlimited usage, with extra resources available. Its founders, Ary and Maria, began the company after encountering deployment problems in their agent-infrastructure research; Maria also taught at MIT, and both worked with Project NANDA. They invite agent developers to share unmet needs, workarounds, and feedback on the product.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.