AI Agent News/stories/maritime-launches-a-managed-cloud-for-production-ai-agents.md

Infrastructure

Maritime launches a managed cloud for production AI agents

Maritime, a Y Combinator F26 startup, has launched a managed cloud that gives each AI agent its own isolated, persistent microVM along with lifecycle management tooling. The company prices a 2GB RAM agent at $1 per month, with unlimited usage included at that price.

Source: news.ycombinator.com/item?id=50549188

The founders, Ary and Maria, say sandboxes solve only part of the infrastructure problem for AI agents. Teams that pick a sandbox provider still have to build sleep and wake, persistent storage, triggers, scheduling, per-user routing, spend caps, health monitoring and zero-downtime updates across many VMs. They also point to demand for custom environments, including Windows and Android, persistent browser sessions, private network access, residential IPs and fully restricted networking.

Maritime gives each agent its own isolated, persistent microVM with a configurable environment. On top of that compute, it offers provisioning, sleep and wake orchestration, triggers, schedules, monitoring, network configuration, rollbacks, recovery and per-agent spend tracking. To control costs it uses scale-to-zero, copy-on-write base images and tiered snapshots, and it mounts a persistent /data volume that can be connected to S3 or R2 buckets.

In the comments, an infrastructure engineer argued that patching hosts and guest kernels after newly found CVEs is harder than launching VMs, and the founders said managed patching is part of the offering. On latency, they said most agent runs take more than 10 seconds, so a few extra milliseconds is rarely a problem, and they plan to add edge hosts and bring-your-own-cloud options.

Why it matters if you build agents

Developers who have built their own lifecycle tooling on top of sandboxes now have a managed option that covers scheduling, recovery, updates and per-agent spend tracking. Those who run agents in distant regions or need fast responses should check where Maritime hosts VMs before committing, since the company has said edge hosts are still planned.

Teams should also ask how quickly the provider patches after vulnerabilities are disclosed, because that ongoing work is the main operational burden of running VMs.

Get stories like this every weekday

One short email with the day's agent tooling news.

Subscribe