Loading post…
CloudflareThomas LefebvreAnnouncement
Cloudflare Containers, rebuilt to scale agent sandboxes
Summary
Cloudflare Containers have been rebuilt to meet the needs of agent sandboxes, which require on-demand creation, fast startup, and the ability to pause and resume. The new durable_object scheduling policy lets application code choose each sandbox’s image and instance type at runtime, replacing the previous deploy-time configuration. A redesigned runtime reduces median container startup from over four seconds to 648 milliseconds, and filesystem snapshots are available in public beta for saving and restoring workspaces. Each container retains its Durable Object as a persistent, programmable controller that manages its lifecycle and outbound traffic. These changes make Containers a more flexible complement to Workers and Durable Objects for agents needing a full Linux workspace.
Context
Agent workloads require on-demand sandboxes that start instantly, can be paused and resumed, and whose image and resources are chosen per task. Previous Cloudflare Containers locked these choices at deploy time, requiring separate applications for each environment. This update makes infrastructure programmable from application code.
Approach / What changed
Cloudflare rearchitected its Containers infrastructure with a new durable_object scheduling policy that moves sandbox configuration (image, instance type) from deploy-time to runtime, a redesigned runtime that starts containers 6x faster (median 648ms), and filesystem snapshots for save/restore. The Durable Object attached to each container now controls its lifecycle directly via the ctx.container API without a wrapper class.
Takeaways
The durable_object scheduling policy lets a single Durable Object class start containers with different images and instance types at runtime, eliminating separate deployments per environment.
Median container startup time dropped from over four seconds to 648 milliseconds in ComputeSDK’s benchmark, enabling near-instant sandbox creation for agent tasks.
Filesystem snapshots (public beta) allow agent workspaces to be saved and restored, supporting long-running projects that span hours or days.
FAQ
How does the new scheduling policy differ from the old one?
The durable_object scheduling policy lets your Durable Object code choose the container image and instance type at runtime via this.ctx.container.start(). Previously, each image/instance combination required a separate application and wrangler deploy.
How much faster do containers start?
In ComputeSDK's independent benchmark, median startup time dropped from just over four seconds to 648 milliseconds. Cloudflare's own burst tests created hundreds of thousands of containers in seconds.
What are filesystem snapshots and how do they work?
Filesystem snapshots are available in public beta. They allow agent workspaces to be saved and restored, so work can continue across sessions without losing the files an agent produced.
Source and freshness
- Source
- blog.cloudflare.com
- Published
- Added to Yomu
Summaries are generated from the original by a language model. Check the original before relying on it.
Keep reading
Related reading
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
The best engineering reads, once a week
One useful email. Unsubscribe anytime.