Google open-sources AX, an agent orchestration system
The pause-and-resume with saved state is the notable part: it converts always-running agent fleets into on-demand workloads, which directly lowers serving costs. Teams building multi-agent backend infrastructure can compare AX against self-managed sandboxing solutions like Firecracker or container-based setups. The GitHub repo lets developers evaluate it for production use immediately.
Google open-sourced AX, a scheduling system for managing and running AI agents at scale.
It gives each agent an isolated virtual sandbox with capped memory and compute to prevent runaway code and interference.
The system pre-provisions code repositories, tool APIs and dependencies before an agent starts working.Agents can be suspended with saved state and resumed later to cut running costs, and administrators can inspect live agent environments for debugging.
The source code is available on GitHub.
AX：谷歌开源的Agent调度系统
专门用来管理和运行“AI 智能体”的调度系统。现在的 AI 不仅能回答问题，还能像真人助理一样，自主使用各种工具、查阅资料甚至编写运行代码来完成复杂任务。但当海量的 AI 助理同时运转时，后台就需要一个强有力的“交通指挥中心”来统一分配计算资源并确保安全。AX 承担的就是这个指挥官的角色。
它主要解决了普通服务器难以应对的几个实际难题：
1）提供安全的独立运行空间：AI 可能会在后台运行各种代码，AX 会给每个 AI 分配独立的虚拟沙盒，限制好内存和算力，防止程序失控或相互干扰。
2）提前准备好工作环境：在 AI 开始工作前，系统会自动把代码库、工具接口和依赖文件准备妥当，让 AI 可以随时投入工作。
3）随时暂停与恢复：当 AI 暂时空闲时，系统能保存当前进度并休眠，等需要时再从原进度继续，以此节省大量的运行成本。
4）支持实时排查：管理者能够随时进入 AI 当前工作的后台环境，像站在 AI 身后一样观察具体进度并排查问题。
Github：https://t.co/gkFbxdSEvx
This report is categorized under AITOP's "Products" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Google, AX, agents, open-source models.
Source attribution: curated from Gorden Sun (@Gorden_Sun).
