Agent Ops: Cloudflare's stack, Google ADK, Kimi K2.6 & CI gates
Building the agentic cloud: everything we launched during Agents Week 2026 launches a suite of agent-first primitives — sandboxes, Git-compatible Artifacts, identity-aware egress, Durable Object Facets, and Workflows v2 — to power an agentic cloud. These primitives change where you draw boundaries for runtimes, egress, and artifact management, giving outcome engineers concrete building blocks for safe, legible landscapes (Principles 06 & 07).
The AI engineering stack we built internally — on the platform we ship describes Cloudflare’s production AI stack that routes billions of tokens and exposes agentic developer tooling like Agents SDK, Sandbox SDK, and Workers AI. It’s a rare, detailed example of platform-first outcome engineering — useful patterns for instrumentation, routing, and running agents at organization scale (Principles 07 & 14).
Orchestrating AI Code Review at scale details a CI-native orchestration that runs specialized AI reviewers, deduplicates findings, and blocks unsafe merges across thousands of PRs. That pattern treats agents as part of an automated delivery pipeline with explicit quality gates and audit hooks — exactly the orchestration and validation posture outcome engineers need (Principles 09 & 15).
Google ADK for Java 1.0 Introduces New App and Plugin Architecture, External Tools Support, and More ships an app/plugin architecture and first-class external tool integrations, plus human-in-the-loop context engineering for agent development. It matters because a supported ADK standardizes how you attach tools, manage context, and implement human handoffs inside production agents — speeding safe, repeatable agent development (Principles 03 & 06).
Kimi K2.6: Advancing Open-Source Coding open-sources K2.6, a coding-focused model that excels at long-horizon, tool-enabled agent workflows and multi-agent orchestration. Having an accessible model tuned for long-horizon orchestration lets outcome engineers iterate on orchestration strategies and reduce dependence on closed APIs for critical delivery pipelines (Principles 03 & 09).