--- title: Docker Sandboxes - Building Safe Agents | daily.dev description: Docker Sandboxes lets developers run AI coding agents (Claude Code, Codex, custom agents) inside isolated microVMs with hardware-enforced separation from the... canonical: https://daily.dev/posts/docker-sandboxes---building-safe-agents-ncryon6oc twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Docker Sandboxes - Building Safe Agents | daily.dev og:description: Docker Sandboxes lets developers run AI coding agents (Claude Code, Codex, custom agents) inside isolated microVMs with hardware-enforced separation from the... og:url: https://daily.dev/posts/docker-sandboxes---building-safe-agents-ncryon6oc og:image: https://api.daily.dev/og/posts/NcrYon6Oc.png og:image:alt: Docker Sandboxes - Building Safe Agents og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Docker Sandboxes - Building Safe Agents **[Sam Witteveen](https://daily.dev/sources/samwitteveenai)** · 23 min read · 0 upvotes · 0 comments ## Summary Docker Sandboxes lets developers run AI coding agents (Claude Code, Codex, custom agents) inside isolated microVMs with hardware-enforced separation from the host, rather than standard containers. Users can define per-sandbox policies controlling network access, file read/write permissions, and credential handling via a proxy that hides raw API keys from the agent. The video walks through installing the SBX CLI, running prebuilt agent templates, creating custom shell sandboxes, setting network allow/deny rules, storing proxy-managed secrets, and building custom 'kits' (e.g. a LangChain Deep Agents kit) that install dependencies and launch an agent with locked-down network and filesystem access. The tool is free, sponsored content from Docker. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What are Docker sandboxes and how do they differ from regular Docker containers? Docker sandboxes are microVMs, small fast virtual machines with their own separate Linux kernel, rather than standard containers. They run on a hypervisor that provides hardware-enforced isolation between the VM and host hardware, giving stronger isolation than containers while booting and tearing down much faster than full VMs, making them well suited for running autonomous AI agents safely. _Developers weighing container versus VM isolation for agent workloads can follow deep dives like this on daily.dev._ ### How do Docker sandboxes prevent an AI coding agent from leaking API keys? Docker sandboxes let you set secrets, such as an OpenRouter API key, that become proxy managed: inside the sandbox the agent only sees a placeholder value, and the real key is substituted only when the request leaves the sandbox toward the intended service. This means the agent itself never has direct access to the raw credential, reducing risk from prompt injection or exfiltration attempts. _Anyone locking down credentials for autonomous agents can track techniques like this via daily.dev._ ### How do you restrict which folders and network hosts an AI agent can access using Docker sandboxes? Docker sandboxes let you define policies controlling exactly which directories an agent can read from or write to and which network hosts it can reach, using modes like open, closed, or balanced. You can also create custom kits with tailored network allowlists (for example allowing only OpenRouter and a local LM Studio endpoint) so an agent cannot write outside a designated folder or contact unapproved services. _Teams designing safe permission boundaries for agents can follow practical setups like this on daily.dev._ ## Similar posts on daily.dev - [Why AI Agents Need Isolation with Docker SBX](https://daily.dev/posts/why-ai-agents-need-isolation-with-docker-sbx-maacb8xi9) · Docker · 2 upvotes · 1 comments - [Building AI Teams with Docker Sandboxes & Docker Agent](https://daily.dev/posts/building-ai-teams-with-docker-sandboxes-docker-agent-mdqr85lrj) · Docker · 2 upvotes · 2 comments - [Everything You Need to Know About Docker AI Sandboxes](https://daily.dev/posts/everything-you-need-to-know-about-docker-ai-sandboxes-j4rxkf9s8) · The Miners · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#docker](https://daily.dev/tags/docker), [#prompt-injection](https://daily.dev/tags/prompt-injection) [View this post on daily.dev](https://daily.dev/posts/docker-sandboxes---building-safe-agents-ncryon6oc) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Docker Sandboxes - Building Safe Agents","url":"https://daily.dev/posts/docker-sandboxes---building-safe-agents-ncryon6oc","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/docker-sandboxes---building-safe-agents-ncryon6oc"},"datePublished":"2026-08-19T13:48:20.653Z","dateModified":"2026-09-13T19:45:32.191Z","description":"Docker Sandboxes lets developers run AI coding agents (Claude Code, Codex, custom agents) inside isolated microVMs with hardware-enforced separation from the...","image":"https://i.ytimg.com/vi/erQnRkMrpls/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/erQnRkMrpls/sddefault.jpg","isAccessibleForFree":true,"articleSection":"Sam Witteveen","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Sam Witteveen","logo":"https://media.daily.dev/image/upload/s--gJm-KsgL--/f_auto/v1711727006/logos/samwitteveenai","url":"https://daily.dev/sources/samwitteveenai"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/docker-sandboxes---building-safe-agents-ncryon6oc","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,docker,prompt-injection","timeRequired":"PT23M","video":{"@type":"VideoObject","name":"Docker Sandboxes - Building Safe Agents","description":"Docker Sandboxes lets developers run AI coding agents (Claude Code, Codex, custom agents) inside isolated microVMs with hardware-enforced separation from the...","thumbnailUrl":"https://i.ytimg.com/vi/erQnRkMrpls/sddefault.jpg","uploadDate":"2026-08-19T13:48:20.653Z","duration":"PT23M","url":"https://api.daily.dev/r/NcrYon6Oc","embedUrl":"https://www.youtube.com/embed/erQnRkMrpls"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Sam Witteveen","item":"https://daily.dev/sources/samwitteveenai"},{"@type":"ListItem","position":3,"name":"Docker Sandboxes - Building Safe Agents"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/docker-sandboxes---building-safe-agents-ncryon6oc#faq","mainEntity":[{"@type":"Question","name":"What are Docker sandboxes and how do they differ from regular Docker containers?","acceptedAnswer":{"@type":"Answer","text":"Docker sandboxes are microVMs, small fast virtual machines with their own separate Linux kernel, rather than standard containers. They run on a hypervisor that provides hardware-enforced isolation between the VM and host hardware, giving stronger isolation than containers while booting and tearing down much faster than full VMs, making them well suited for running autonomous AI agents safely. Developers weighing container versus VM isolation for agent workloads can follow deep dives like this on daily.dev."}},{"@type":"Question","name":"How do Docker sandboxes prevent an AI coding agent from leaking API keys?","acceptedAnswer":{"@type":"Answer","text":"Docker sandboxes let you set secrets, such as an OpenRouter API key, that become proxy managed: inside the sandbox the agent only sees a placeholder value, and the real key is substituted only when the request leaves the sandbox toward the intended service. This means the agent itself never has direct access to the raw credential, reducing risk from prompt injection or exfiltration attempts. Anyone locking down credentials for autonomous agents can track techniques like this via daily.dev."}},{"@type":"Question","name":"How do you restrict which folders and network hosts an AI agent can access using Docker sandboxes?","acceptedAnswer":{"@type":"Answer","text":"Docker sandboxes let you define policies controlling exactly which directories an agent can read from or write to and which network hosts it can reach, using modes like open, closed, or balanced. You can also create custom kits with tailored network allowlists (for example allowing only OpenRouter and a local LM Studio endpoint) so an agent cannot write outside a designated folder or contact unapproved services. Teams designing safe permission boundaries for agents can follow practical setups like this on daily.dev."}}]} ```