WRITER launched its platform in 2020 with early transformer models powering text generation and content creation. But as the models and supporting infrastructure evolved to support complex workflows, customers started putting agents to work on more sophisticated, multi-step enterprise tasks. To meet this growing demand, WRITER needed isolated, scalable sandbox environments where agents could operate safely and reliably.

Initially, WRITER built an in-house sandbox platform on AWS Fargate to power its multi-agent orchestration graph. This infrastructure dynamically provisioned and terminated dedicated sandboxes for a set of agent tasks, minimizing latency to protect the user experience and control infrastructure costs. However, maintaining this performance at a scale of 20,000+ sandboxes per day introduced immense operational overhead.

“Provisioning compute capacity and maintaining reliability meant our team was spending days managing sandboxes, rather than advancing our models and agent capabilities,” shares Waseem Alshikh, Co-Founder and CTO at WRITER.

That overhead wasn’t limited to scale. Supporting enterprise customers required robust security and governance to protect sensitive data and meet evolving requirements for data residency, access controls, and auditability. Each sandbox had to securely execute untrusted code, so strong workload isolation was also non-negotiable.

For WRITER’s engineering team, building and maintaining infrastructure to meet all of those requirements wasn’t the best use of their time. Instead, Waseem began evaluating sandbox infrastructure providers that could support WRITER’s agent workloads in production. While several solutions seemed promising, most required bespoke architectural setups to support throughput increases, creating significant friction.

Fortunately, it wasn’t long before Waseem discovered Daytona. Their sandbox infrastructure met every requirement to support WRITER’s growing agent workloads and keep engineering focused on advancing its AI platform.