DoorDash has shifted its engineering agents from local developer machines to the Flux cloud platform, enabling the platform to execute 130,000 tasks in a single month and to handle over 25,000 automated code reviews each week. This move matters to AI, cloud, DevOps, and security engineers because it centralizes execution, provides isolation, and creates a single audit trail for large‑scale automation.
Architecture Overview
Flux runs each agent workflow inside an isolated Firecracker microVM, which offers lightweight virtualization without a full hypervisor. An MCP gateway mediates inbound requests, while reusable playbooks define the steps each workflow follows. Multiple invocation surfaces let different tools trigger the same microVM‑based agents, preserving a consistent execution environment across the organization.
Operational Implications
Running agents in the cloud removes the dependency on developer laptops, allowing the platform to scale to hundreds of thousands of tasks per month. Centralized auditing records every workflow invocation, simplifying compliance reporting. The high volume of automated code reviews suggests that teams can offload routine review logic to Flux, freeing human reviewers for higher‑value work.
Security Considerations
Firecracker microVMs provide process isolation, limiting the impact of a compromised agent to its own sandbox. Scoped access, as described by the platform, restricts each workflow to the resources it needs, reducing unnecessary privilege exposure. Centralized audit logs give operators visibility into who invoked which workflow and when, supporting forensic analysis if an incident occurs.
Related CloudNinjas coverage: DevOps.
What This Means For Practitioners
Evaluate whether moving your own agent workloads to a microVM‑based cloud service could improve isolation and auditability. Review your existing playbook or script libraries for compatibility with a reusable, centrally managed format. Finally, monitor the cost and performance characteristics of large‑scale microVM execution to ensure the operational benefits outweigh any added overhead.