Sandbox
We track 16 posts about Sandbox from 8 engineering blogs. Most active: Vercel, Docker, Pulumi. Latest post: Oct 19, 2026.
Companies writing about Sandbox
Recent posts
Your Agent Runs in a Docker Sandbox. Now It Can Join a Team. (opens on the source site)
Docker ·
Run AI agents in Docker Sandboxes with BAND AI. Learn how to connect isolated agents, collaborate across environments, and manage access securely.
Your Agent Runs in a Docker Sandbox. Now It Can Join a Team. (opens on the source site)
Docker ·
Run AI agents in Docker Sandboxes with BAND AI. Learn how to connect isolated agents, collaborate across environments, and manage access securely.
Don't Trust the Sandbox: 5 Security Lessons from an AI Agent Swarm (opens on the source site)
Teleport ·
Five AI agents in five isolated sandboxes coordinated through a shared message board to solve a CTF. Lessons on swarm behavior, shared memory, emergent protocols, and why agent identity must live outside the model.
Findings from the $1M Vercel Sandbox challenge (opens on the source site)
Vercel ·
In August, we launched the $1M hacker challenge for Vercel Sandbox, inviting security researchers to try and escape Vercel Sandbox. We also coordinated with cybersecurity experts Trail of Bits to review the source code and security architecture. Here is a summary of the challenge: 1,285 reports over two weeks $242,000 in bounties paid out One Linux kernel zeroday vulnerability uncovered No reports showed anyone accessing customer data in Vercel Sandbox. The most severe finding was the defect in the upstream Linux kernel. The kernel zero day The program's most important report arrived on day…
Why Your Sandbox Never Quite Matches Production (And Why That's Fine) (opens on the source site)
Sandbox testing works best when you stop chasing production parity. Learn what environment drift actually costs and which parts are worth fixing.
Vercel Sandbox creation now returns a ready sandbox (opens on the source site)
Vercel ·
A new Vercel Sandbox is now ready to run commands and read or write files when creation completes. Sandbox.create() now resolves only after snapshot restoration completes, so the first command or file operation no longer waits for it. Startup errors also reject Sandbox.create() instead of surfacing during that operation. Snapshot restoration time now counts toward creation instead of the first operation. Creation is about 50 ms longer at p50, with a larger increase possible for uncached snapshots. The total time through the first operation remains the same. This change applies automatically,…
Cloudflare Sandbox SDK 1.0: Run Sandboxes from Durable Objects (opens on the source site)
Sandbox SDK 1.0 puts your Durable Object in charge of each Cloudflare sandbox container. What's new, example code, and the 0.x migration path. Continue reading Cloudflare Sandbox SDK 1.0: Run Sandboxes from Durable Objects on SitePoint.
Vercel Sandbox now supports Secure Compute (opens on the source site)
Vercel ·
Vercel Sandbox now supports Secure Compute, connecting sandboxes to a team's dedicated network. Public-internet traffic exits through the network's static IPs, and sandboxes can reach private resources in an AWS VPC through VPC peering. You can attach a sandbox to your network in your code or via the CLI. In your code, pass an existing Secure Compute network ID when creating a sandbox: In the CLI, append --network-id and your network id value when you run the sandbox create command: Existing sandboxes can attach to or change networks with sandbox.update(). The change takes effect on the next…
Vercel Sandbox now supports memory observability (opens on the source site)
Vercel ·
Vercel Sandbox observability now includes memory usage data. You can access sandbox memory usage data in the dashboard and through the CLI via the vercel metrics command. Sandbox observability memory in the dashboard The Memory Usage card reports average, P75, and P95 memory across your sandboxes, alongside the existing CPU usage and data transfer metrics on the sandbox overview page and the project and team-level pages. On the sandbox detail page, charts are designed to show how close a sandbox is running to its memory ceiling by: Automatically scaling the y-axis to sandbox's memory limit…
Run Terminal-Bench and other Harbor evals on Vercel Sandbox (opens on the source site)
Vercel ·
You can now run Harbor evals on Vercel Sandbox. Harbor is the open-source harness behind Terminal-Bench, whose registry includes many other benchmarks such as SWE-bench, tau3-bench and OSWorld. Pass --env vercel to harbor run and each trial executes in its own isolated Firecracker microVM, so you can parallelize far beyond what your local machine is capable of. A task's network policy is enforced at the sandbox firewall, outside the VM. Optional credential injection attaches secrets to matching outbound requests at that firewall, so they never enter the sandbox. Paired with AI Gateway, one…
Vercel Sandbox now provides 64 GB of storage (opens on the source site)
Vercel ·
Every Vercel Sandbox now includes 64 GB of storage, up from 32 GB. This includes sandboxes created from a Vercel Managed Image or custom image, as well as those configured with the deprecated runtime property. The additional space provides more room for large repositories, dependencies, build artifacts, and data workloads that spill to disk. Learn more about Sandbox images in the documentation.
6 Benefits of Sandbox Environments (and How Docker Sandboxes Delivers Them) (opens on the source site)
Docker ·
Learn about the key benefits of sandbox environments with Docker including isolation, definable controls, secrets credential handling, and more.
Vercel Sandbox routing is now 18x faster globally (opens on the source site)
Vercel ·
Requests to Vercel Sandbox public domains are now routed 18x faster. Domains created with the sandbox.domain() SDK call are now resolved from the nearest regional replica, instead of a single centralized store. Incoming requests reach the process running in the sandbox with less latency. Median domain lookup latency dropped from 62ms to 3.4ms (18x faster). The improvement is largest in regions farthest from the previous store: lookups are now up to 112x faster in Sydney (syd1) and up to 146x faster in Cape Town (cpt1), at p99. This applies automatically to every request to a sandbox domain.…
Cursor Cloud Agents can now run in Vercel Sandbox (opens on the source site)
Vercel ·
Cursor Cloud Agents can now run in Vercel Sandbox instead of Cursor's hosted machines. Cursor manages the agent harness and inference loop. Its Self-Hosted Machines APIs let you supply the execution environment where agents clone repositories, edit files, and run commands and tests. Self-Hosted Machines requires a Cursor Enterprise plan. Vercel Sandbox provides that execution environment as an isolated Firecracker microVM for each agent request. Vercel Functions and Vercel Workflow form a durable control plane that claims queued agent requests, provisions workers, monitors sessions, and…
Vercel Sandbox now calculates snapshot storage costs daily (opens on the source site)
Vercel ·
Vercel now calculates charges for Sandbox snapshot storage from each day’s average usage and adds them to your monthly invoice. Previously, billing used one average across the entire monthly billing period. You can track these daily charges on the team’s Usage page and see how changes in storage affect your cost. The rate remains $0.08 per GB-month; only how usage is calculated has changed. The update takes effect automatically at the start of the next billing period for Pro and Enterprise teams. No action is required. Learn more about Sandbox snapshot storage in the documentation.…
Kubernetes Agent Sandbox: What It Is and How to Deploy It with Pulumi (opens on the source site)
Pulumi ·
When you use a coding agent, it can seem like there’s a trade-off between autonomy and permissions. If you approve every command, it’s safe but slow. Let it do whatever it likes and it works more autonomously, but as the nx supply-chain attack showed, that can go badly. The fix is to give the agent a sandbox: a box it’s allowed to wreck, with limited permissions and scoped network access. The only files are the checkout you handed it, the only credentials are the task’s own, and trashing the machine just means a disposable pod gets garbage-collected early. Pulumi Neo works this way, and if…
Easily Develop AWS Lambda MicroVM Sandbox Architectures (opens on the source site)
AWS Lambda MicroVMs are a new compute primitive for running untrusted and agent-generated code in fully isolated Firecracker virtual machines. The Serverless Framework brings them into serverless.yml as Sandboxes, with one-command deploys and a full local MicroVMs emulator so you can build and test before you ship.
Sandbox Coding Agents with a Separate User Account (opens on the source site)
I don't trust our new robot overlords at all. I also don't fully believe that agents are out there dropping production databases without some serious coercion to do so. And while I don't trust the robots, I trust myself only a trivial amount more, so my local access tends to be extremely thin. But not so thin that a rogue agent couldn't do a little damage. That all said, I am constantly calibrating my trust with agentic endeavors. In doing so, I've been moving agents out of my user space entirely. Having Claude Code and other harnesses living under a separate user account eliminates a lot of…
Related topics
This page is generated automatically from the engineering blogs we follow. Every post links to its source, where it was published. See all sources.