# Agent Sandboxes > Independent, source-cited comparison of AI agent sandboxes: E2B, Daytona, Modal, Vercel Sandbox, Cloudflare, Fly.io Sprites and more. Isolation, cold starts, persistence, pricing and which sandbox fits your agent. Agent Sandboxes (agentsandboxes.org) is an independent, source-cited reference on AI agent sandboxes — isolated environments where AI agents run code. It profiles 21 active providers and open-source projects. No provider pays for placement. Facts are attributed to provider documentation and dated; cold-start figures are vendor claims. Last reviewed: 2026-11-08. Quick answer: The best AI agent sandboxes in 2026 are E2B (the most widely adopted, open-source microVM sandbox), Daytona and Modal Sandboxes. The right choice depends on your workload: coding agents → Daytona; code interpreters → E2B; RL training and evals → Modal Sandboxes; computer-use agents → Daytona. ## Core pages - [Best AI agent sandboxes in 2026](https://agentsandboxes.org/): Independent ranking of the best AI agent sandboxes in 2026: E2B, Daytona, Modal Sandboxes and 18 more compared on isolation, cold start, persistence, GPUs and pricing. Sourced and dated. - [Agent sandbox directory](https://agentsandboxes.org/sandboxes/): Directory of 21 AI agent sandbox providers and open-source projects — hosted platforms, cloud services and self-hostable tools — with isolation, deployment and use-case notes. - [Compare agent sandboxes](https://agentsandboxes.org/compare/): Compare 21 AI agent sandboxes side by side: isolation technology, open source, self-hosting, cold start, max session, snapshots, GPUs and SDKs. Filterable and sourced. - [Best agent sandbox by use case](https://agentsandboxes.org/best/): The best AI agent sandboxes ranked for each workload: coding agents, code interpreters, computer use, RL and evals, GPUs, self-hosting, enterprise and local agents. - [Methodology: how we evaluate agent sandboxes](https://agentsandboxes.org/methodology/): How agentsandboxes.org selects, documents and ranks AI agent sandbox providers: sources, criteria, use-case weighting, update cadence and corrections. ## Best sandbox by use case - [Best sandboxes for coding agents](https://agentsandboxes.org/best/coding-agents/): The best AI agent sandboxes for coding agents in 2026 — persistent workspaces, fast resume, Git, preview URLs and long sessions — ranked with sources. - [Best sandboxes for code interpreters and data analysis](https://agentsandboxes.org/best/code-interpreter/): The best sandboxes for adding a code interpreter to an AI app — stateful Python, charts, file I/O, fast startup and strong isolation — compared and ranked. - [Best sandboxes for reinforcement learning and agent evals](https://agentsandboxes.org/best/rl-and-evals/): The best sandboxes for agentic reinforcement learning and evaluations — massive concurrency, snapshot forking, fast reset and per-second billing — ranked. - [Best sandboxes for computer-use and browser agents](https://agentsandboxes.org/best/computer-use/): The best sandboxes for computer-use and browser agents — virtual desktops, VNC/streaming, Windows and macOS options and long sessions — compared. - [Best GPU sandboxes for AI agents](https://agentsandboxes.org/best/gpu/): Agent sandboxes with GPU support compared — Modal, Daytona, Northflank, Beam and more — GPU types, isolation, pricing and limits. - [Best open-source and self-hosted agent sandboxes](https://agentsandboxes.org/best/open-source/): The best open-source agent sandboxes you can self-host — E2B, Kubernetes Agent Sandbox, Microsandbox and more — licence, isolation and operational effort compared. - [Best agent sandboxes for enterprise and BYOC](https://agentsandboxes.org/best/enterprise/): The best agent sandboxes for enterprises — BYOC in your own cloud, compliance (SOC 2, HIPAA), VPC integration and data residency — compared. - [Best sandboxes for local coding agents](https://agentsandboxes.org/best/local-agents/): The best ways to sandbox AI coding agents on your own laptop — Docker Sandboxes, Anthropic Sandbox Runtime and Microsandbox — isolation and trade-offs compared. ## Provider profiles - [E2B](https://agentsandboxes.org/sandboxes/e2b/): E2B agent sandbox: Open-source cloud sandboxes for AI agents, each a Firecracker microVM, with pause/resume, snapshots and forking. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Daytona](https://agentsandboxes.org/sandboxes/daytona/): Daytona agent sandbox: Stateful sandboxes for AI-generated code — Linux containers, Linux and Windows VMs, macOS and GPU sandboxes — with snapshots, volumes and preview URLs. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Modal Sandboxes](https://agentsandboxes.org/sandboxes/modal/): Modal Sandboxes agent sandbox: Sandboxes on Modal’s serverless compute platform for running untrusted code at very high scale, with GPUs, snapshots and tunnels. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Vercel Sandbox](https://agentsandboxes.org/sandboxes/vercel-sandbox/): Vercel Sandbox agent sandbox: Firecracker microVM sandboxes on Vercel’s infrastructure, persistent by default and billed on active CPU, with fine-grained egress control. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Cloudflare Sandbox SDK](https://agentsandboxes.org/sandboxes/cloudflare-sandbox/): Cloudflare Sandbox SDK agent sandbox: A TypeScript SDK for Cloudflare Workers that runs untrusted code in Linux sandboxes backed by Cloudflare Containers and Durable Objects. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Fly.io Sprites](https://agentsandboxes.org/sandboxes/fly-sprites/): Fly.io Sprites agent sandbox: Persistent, hardware-isolated Linux computers for AI agents with a durable 100 GB disk, checkpoints, and no compute charges while idle. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Northflank Sandboxes](https://agentsandboxes.org/sandboxes/northflank/): Northflank Sandboxes agent sandbox: Sandboxes on Northflank’s Kubernetes-based platform — microVMs for CPU work, gVisor for GPUs — on Northflank’s cloud or in your own cloud account. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Blaxel](https://agentsandboxes.org/sandboxes/blaxel/): Blaxel agent sandbox: Perpetual microVM sandboxes for AI agents that drop to standby when idle and resume with memory intact, plus volumes, batch jobs and MCP hosting. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Together Code Sandbox](https://agentsandboxes.org/sandboxes/codesandbox/): Together Code Sandbox agent sandbox: Together AI’s microVM sandbox service, built on the CodeSandbox acquisition, with snapshot-based starts, hibernation, cloning and preview URLs. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Morph Cloud](https://agentsandboxes.org/sandboxes/morph/): Morph Cloud agent sandbox: VM-based cloud computers for AI agents built on “Infinibranch” snapshot, branch and restore, plus persistent Devboxes. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Beam Sandboxes](https://agentsandboxes.org/sandboxes/beam/): Beam Sandboxes agent sandbox: Serverless AI compute platform whose sandboxes are gVisor-isolated containers with memory snapshots, optional GPUs and an open-source, self-hostable runtime. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Freestyle](https://agentsandboxes.org/sandboxes/freestyle/): Freestyle agent sandbox: Full Linux VMs for AI agents with live forking of running VMs, memory-preserving hibernation, VPCs and nested virtualization. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Koyeb Sandboxes](https://agentsandboxes.org/sandboxes/koyeb/): Koyeb Sandboxes agent sandbox: Ephemeral microVM sandboxes on Koyeb’s serverless platform for AI-generated code, with CPU or GPU and scale-to-zero. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Runloop (winding down)](https://agentsandboxes.org/sandboxes/runloop/): Runloop agent sandbox: Managed microVM “Devboxes” for AI coding agents with blueprints, snapshots, suspend/resume and benchmarking tools — now winding down. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Riza (discontinued)](https://agentsandboxes.org/sandboxes/riza/): Riza agent sandbox: A former hosted code-interpreter API for running LLM-generated code; the hosted service shut down in October 2025. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [AWS Bedrock AgentCore](https://agentsandboxes.org/sandboxes/aws-agentcore/): AWS Bedrock AgentCore agent sandbox: AWS’s managed agent platform: a Code Interpreter tool for running agent-generated code and a Runtime that hosts each agent session in its own microVM. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Azure Container Apps dynamic sessions](https://agentsandboxes.org/sandboxes/azure-dynamic-sessions/): Azure Container Apps dynamic sessions agent sandbox: Azure service that hands out Hyper-V-isolated sessions from prewarmed pools, either as a built-in code interpreter or running your own container image. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 20 - [Google Agent Platform Code Execution](https://agentsandboxes.org/sandboxes/google-code-execution/): Google Agent Platform Code Execution agent sandbox: Google-managed sandbox API (formerly part of Vertex AI Agent Engine) that lets agents deployed anywhere run Python in stateful, network-isolated environments. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November - [OpenAI Code Interpreter & Hosted Shell](https://agentsandboxes.org/sandboxes/openai-code-interpreter/): OpenAI Code Interpreter & Hosted Shell agent sandbox: Hosted containers in the OpenAI Responses API that the model uses to run Python (Code Interpreter) or shell commands (Hosted Shell). Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Kubernetes Agent Sandbox](https://agentsandboxes.org/sandboxes/gke-agent-sandbox/): Kubernetes Agent Sandbox agent sandbox: Open-source Kubernetes controller and CRDs for isolated, stateful sandbox pods for AI agents, with first-class support on GKE. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Microsandbox](https://agentsandboxes.org/sandboxes/microsandbox/): Microsandbox agent sandbox: Open-source, local-first microVM runtime that gives each sandbox its own Linux kernel, with an optional managed cloud. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Docker Sandboxes](https://agentsandboxes.org/sandboxes/docker-sandboxes/): Docker Sandboxes agent sandbox: Docker’s `sbx` CLI for running AI coding agents in microVM sandboxes on your machine, with an experimental Docker-hosted cloud. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. - [Anthropic Sandbox Runtime](https://agentsandboxes.org/sandboxes/anthropic-srt/): Anthropic Sandbox Runtime agent sandbox: Open-source CLI and library (`srt`) that restricts filesystem and network access for any local process at the OS level, without a container or VM. Isolation, cold start, persistence, GPUs, pricing and how it compares — verified November 8, 2026. ## Head-to-head comparisons - [E2B vs Daytona](https://agentsandboxes.org/compare/e2b-vs-daytona/): E2B vs Daytona compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [E2B vs Modal Sandboxes](https://agentsandboxes.org/compare/e2b-vs-modal/): E2B vs Modal Sandboxes compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [E2B vs Vercel Sandbox](https://agentsandboxes.org/compare/e2b-vs-vercel-sandbox/): E2B vs Vercel Sandbox compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [E2B vs Cloudflare Sandbox SDK](https://agentsandboxes.org/compare/e2b-vs-cloudflare-sandbox/): E2B vs Cloudflare Sandbox SDK compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [E2B vs Fly.io Sprites](https://agentsandboxes.org/compare/e2b-vs-fly-sprites/): E2B vs Fly.io Sprites compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Daytona vs Modal Sandboxes](https://agentsandboxes.org/compare/daytona-vs-modal/): Daytona vs Modal Sandboxes compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Daytona vs Vercel Sandbox](https://agentsandboxes.org/compare/daytona-vs-vercel-sandbox/): Daytona vs Vercel Sandbox compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Daytona vs Cloudflare Sandbox SDK](https://agentsandboxes.org/compare/daytona-vs-cloudflare-sandbox/): Daytona vs Cloudflare Sandbox SDK compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Daytona vs Fly.io Sprites](https://agentsandboxes.org/compare/daytona-vs-fly-sprites/): Daytona vs Fly.io Sprites compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Modal Sandboxes vs Vercel Sandbox](https://agentsandboxes.org/compare/modal-vs-vercel-sandbox/): Modal Sandboxes vs Vercel Sandbox compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Modal Sandboxes vs Cloudflare Sandbox SDK](https://agentsandboxes.org/compare/modal-vs-cloudflare-sandbox/): Modal Sandboxes vs Cloudflare Sandbox SDK compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Modal Sandboxes vs Fly.io Sprites](https://agentsandboxes.org/compare/modal-vs-fly-sprites/): Modal Sandboxes vs Fly.io Sprites compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Vercel Sandbox vs Cloudflare Sandbox SDK](https://agentsandboxes.org/compare/vercel-sandbox-vs-cloudflare-sandbox/): Vercel Sandbox vs Cloudflare Sandbox SDK compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Vercel Sandbox vs Fly.io Sprites](https://agentsandboxes.org/compare/vercel-sandbox-vs-fly-sprites/): Vercel Sandbox vs Fly.io Sprites compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Cloudflare Sandbox SDK vs Fly.io Sprites](https://agentsandboxes.org/compare/cloudflare-sandbox-vs-fly-sprites/): Cloudflare Sandbox SDK vs Fly.io Sprites compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Northflank Sandboxes vs Modal Sandboxes](https://agentsandboxes.org/compare/northflank-vs-modal/): Northflank Sandboxes vs Modal Sandboxes compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Northflank Sandboxes vs E2B](https://agentsandboxes.org/compare/northflank-vs-e2b/): Northflank Sandboxes vs E2B compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Blaxel vs E2B](https://agentsandboxes.org/compare/blaxel-vs-e2b/): Blaxel vs E2B compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Blaxel vs Fly.io Sprites](https://agentsandboxes.org/compare/blaxel-vs-fly-sprites/): Blaxel vs Fly.io Sprites compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Together Code Sandbox vs E2B](https://agentsandboxes.org/compare/codesandbox-vs-e2b/): Together Code Sandbox vs E2B compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Morph Cloud vs E2B](https://agentsandboxes.org/compare/morph-vs-e2b/): Morph Cloud vs E2B compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Beam Sandboxes vs Modal Sandboxes](https://agentsandboxes.org/compare/beam-vs-modal/): Beam Sandboxes vs Modal Sandboxes compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Freestyle vs Daytona](https://agentsandboxes.org/compare/freestyle-vs-daytona/): Freestyle vs Daytona compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Freestyle vs E2B](https://agentsandboxes.org/compare/freestyle-vs-e2b/): Freestyle vs E2B compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [AWS Bedrock AgentCore vs E2B](https://agentsandboxes.org/compare/aws-agentcore-vs-e2b/): AWS Bedrock AgentCore vs E2B compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Microsandbox vs E2B](https://agentsandboxes.org/compare/microsandbox-vs-e2b/): Microsandbox vs E2B compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. - [Docker Sandboxes vs Anthropic Sandbox Runtime](https://agentsandboxes.org/compare/docker-sandboxes-vs-anthropic-srt/): Docker Sandboxes vs Anthropic Sandbox Runtime compared for AI agents: isolation, cold start, persistence, GPUs, pricing, open source and which to choose. Sourced and dated. ## Alternatives - [Best E2B alternatives](https://agentsandboxes.org/alternatives/e2b/): The best E2B alternatives for AI agent sandboxes in 2026: Daytona, Modal Sandboxes, Vercel Sandbox, Fly.io Sprites, Cloudflare Sandbox SDK and more — compared on isolation, persistence, GPUs, pricing and open source. - [Best Daytona alternatives](https://agentsandboxes.org/alternatives/daytona/): The best Daytona alternatives for AI agent sandboxes in 2026: E2B, Fly.io Sprites, Modal Sandboxes, Vercel Sandbox, Together Code Sandbox and more — compared on isolation, persistence, GPUs, pricing and open source. - [Best Modal Sandboxes alternatives](https://agentsandboxes.org/alternatives/modal/): The best Modal Sandboxes alternatives for AI agent sandboxes in 2026: E2B, Daytona, Beam Sandboxes, Northflank Sandboxes, Morph Cloud and more — compared on isolation, persistence, GPUs, pricing and open source. - [Best Vercel Sandbox alternatives](https://agentsandboxes.org/alternatives/vercel-sandbox/): The best Vercel Sandbox alternatives for AI agent sandboxes in 2026: E2B, Cloudflare Sandbox SDK, Daytona, Fly.io Sprites, Freestyle and more — compared on isolation, persistence, GPUs, pricing and open source. - [Best Cloudflare Sandbox SDK alternatives](https://agentsandboxes.org/alternatives/cloudflare-sandbox/): The best Cloudflare Sandbox SDK alternatives for AI agent sandboxes in 2026: Vercel Sandbox, E2B, Fly.io Sprites, Freestyle, Daytona and more — compared on isolation, persistence, GPUs, pricing and open source. - [Best Fly.io Sprites alternatives](https://agentsandboxes.org/alternatives/fly-sprites/): The best Fly.io Sprites alternatives for AI agent sandboxes in 2026: Daytona, E2B, Vercel Sandbox, Morph Cloud, Together Code Sandbox and more — compared on isolation, persistence, GPUs, pricing and open source. - [Best Runloop alternatives](https://agentsandboxes.org/alternatives/runloop/): The best Runloop alternatives for AI agent sandboxes in 2026: Freestyle, Modal Sandboxes, Daytona, E2B, Fly.io Sprites and more — compared on isolation, persistence, GPUs, pricing and open source. - [Best Riza alternatives](https://agentsandboxes.org/alternatives/riza/): The best Riza alternatives for AI agent sandboxes in 2026: E2B, Daytona, Azure Container Apps dynamic sessions, AWS Bedrock AgentCore, OpenAI Code Interpreter & Hosted Shell and more — compared on isolation, persistence, GPUs, pricing and open source. ## Guides - [What is an AI agent sandbox?](https://agentsandboxes.org/guides/what-is-an-agent-sandbox/): A plain-English explanation of AI agent sandboxes: what they are, why agents need them, how they are isolated and what to look for. - [MicroVMs vs gVisor vs containers for AI agents](https://agentsandboxes.org/guides/microvm-vs-gvisor-vs-containers/): Firecracker microVMs, gVisor, Kata Containers, plain containers and V8 isolates compared for running untrusted AI-generated code: security boundary, startup, overhead and compatibility. - [How to choose an agent sandbox: a buyer’s checklist](https://agentsandboxes.org/guides/how-to-choose-an-agent-sandbox/): A practical checklist for choosing an AI agent sandbox provider: workload fit, isolation, cold start, persistence, networking, pricing, SDKs, compliance and lock-in. - [Agent sandbox security checklist](https://agentsandboxes.org/guides/agent-sandbox-security-checklist/): A security checklist for running AI agents in sandboxes: isolation, network egress, secrets, resource limits, prompt injection, auditing and human approval. - [Agent sandbox pricing explained](https://agentsandboxes.org/guides/agent-sandbox-pricing/): How AI agent sandbox pricing works — per-second vCPU and memory billing, free tiers, storage, idle time — with a worked example and links to each provider’s published rates. - [Self-hosting agent sandboxes: open-source options](https://agentsandboxes.org/guides/self-hosting-agent-sandboxes/): How to self-host AI agent sandboxes with open-source software — E2B, Daytona, Microsandbox, Kubernetes agent-sandbox, Kata Containers, gVisor and Firecracker — and what it takes to run them. ## Optional - [Full text for LLMs](https://agentsandboxes.org/llms-full.txt): every provider record and guide in one plain-text file - [Open data (JSON)](https://agentsandboxes.org/data/providers.json): structured provider records, CC BY 4.0 - [Agent sandbox](https://agentsandboxes.org/glossary/agent-sandbox/): An isolated, disposable compute environment where an AI agent can run code, use tools and touch files without risking the host or other users. - [MicroVM](https://agentsandboxes.org/glossary/microvm/): A minimal virtual machine with a stripped-down device model that boots in milliseconds and gives each workload its own guest kernel. - [Firecracker](https://agentsandboxes.org/glossary/firecracker/): An open-source virtual machine monitor from AWS, built on KVM, designed to run thousands of lightweight microVMs per host. - [gVisor](https://agentsandboxes.org/glossary/gvisor/): An open-source application kernel from Google that intercepts a container’s system calls in user space, adding a boundary between the workload and the host kernel. - [Kata Containers](https://agentsandboxes.org/glossary/kata-containers/): An open-source runtime that runs each container (or pod) inside its own lightweight VM while keeping the standard container interface. - [Cloud Hypervisor](https://agentsandboxes.org/glossary/cloud-hypervisor/): An open-source, Rust-based virtual machine monitor for modern cloud workloads, governed under the Linux Foundation. - [Container isolation](https://agentsandboxes.org/glossary/container-isolation/): Isolation built from Linux namespaces, cgroups, capabilities and seccomp, where workloads share the host kernel. - [seccomp](https://agentsandboxes.org/glossary/seccomp/): A Linux kernel feature that restricts which system calls a process may make. - [V8 isolate](https://agentsandboxes.org/glossary/v8-isolate/): A lightweight execution context inside the V8 JavaScript engine; many isolates can share one process with separate memory heaps. - [WebAssembly sandbox](https://agentsandboxes.org/glossary/wasm-sandbox/): Running code compiled to WebAssembly in a runtime that only grants the capabilities you pass in. - [Cold start](https://agentsandboxes.org/glossary/cold-start/): The time from requesting a new sandbox to being able to run the first command in it. - [Snapshot](https://agentsandboxes.org/glossary/snapshot/): A saved copy of a sandbox’s filesystem — and sometimes its memory and running processes — that can be restored later. - [Pause and resume](https://agentsandboxes.org/glossary/pause-resume/): Suspending a sandbox so it stops consuming compute, then restoring it later with its state intact. - [Fork (branching)](https://agentsandboxes.org/glossary/fork/): Creating one or more copies of a running sandbox from a snapshot so several agent attempts can proceed in parallel from the same state. - [Persistent (stateful) sandbox](https://agentsandboxes.org/glossary/persistent-sandbox/): A sandbox whose files and configuration survive beyond a single session, often for days or indefinitely. - [Ephemeral sandbox](https://agentsandboxes.org/glossary/ephemeral-sandbox/): A short-lived sandbox created for one task and destroyed afterwards, leaving no state behind. - [Egress control](https://agentsandboxes.org/glossary/egress-control/): Rules that limit which network destinations code inside a sandbox can reach. - [Prompt injection](https://agentsandboxes.org/glossary/prompt-injection/): An attack where instructions hidden in content an agent reads (a web page, file or issue) hijack what the agent does. - [Secrets handling](https://agentsandboxes.org/glossary/secrets-handling/): How API keys and credentials are made available to code in a sandbox without being exposed to the model or leaked. - [BYOC (bring your own cloud)](https://agentsandboxes.org/glossary/byoc/): A deployment model where the provider’s sandbox platform runs inside your own cloud account or VPC. - [Self-hosted sandbox](https://agentsandboxes.org/glossary/self-hosted-sandbox/): Running sandbox infrastructure on servers or clusters you operate, typically using open-source software. - [Code interpreter](https://agentsandboxes.org/glossary/code-interpreter/): A tool that lets a model write and execute code (usually Python) and see the output, typically for analysis, math and file conversion. - [Computer use](https://agentsandboxes.org/glossary/computer-use/): Agents that operate a graphical desktop or browser through screenshots, mouse and keyboard actions. - [MCP (Model Context Protocol)](https://agentsandboxes.org/glossary/mcp/): An open protocol for connecting AI applications to tools and data sources through a standard client–server interface. - [RL environment](https://agentsandboxes.org/glossary/rl-environment/): A reproducible, resettable environment in which an agent takes actions and receives rewards, used to train or evaluate models.