Agent infrastructure: runtimes, orchestration, and security

Big Cloud™ touts swappable AI models, wants the harness. Big Cloud pitches swappable models and a harness to route requests around model refusals, expanding design space for resilient AI services. This matters because model-routing and harnesses let outcome engineers build failover, A/Bing, and safety fences at the request layer — think resilient delivery lanes, not brittle single-model stacks (Principles 09 & 11).

June emerges from stealth to automate enterprise AI deployment with agents and bottleneck detection. June launches with $20M to detect deployment bottlenecks and spin up task agents that automate rollout steps. Outcome engineers gain an operational pattern for automating the last mile of deployment and surfacing rollout chokepoints early, reducing manual handoffs (Principles 04 & 09).

Your agent needs a computer, not a container — introducing @cloudflare/computer. Cloudflare open-sources an isolate-first agent runtime that gives each agent a fast, secure compute environment instead of heavyweight containers. That directly affects how you design agent runtimes and scaling: lighter isolates mean denser, cheaper agent fleets and clearer boundaries for auditing and containment (Principles 07 & 12).

Orchard: An open framework for scalable agentic AI. Microsoft releases Orchard with Kubernetes-based environments and training recipes to build and evaluate agentic systems across domains. Outcome engineers get a repeatable testbed for measuring agent behavior, running staged experiments, and integrating environment-level controls into development workflows (Principles 06 & 07).

Zenity raises $125M Series C to secure AI agents. Zenity secures $125M to accelerate enterprise tooling that enforces real-time security and governance for autonomous agents. For outcome engineering teams, that signals mature vendor tooling for runtime monitoring, policy enforcement, and mitigation — the operational controls you need before agents run unsupervised in production (Principles 14 & 15).