--- title: "Guides to microVMs, Sandboxes, and Cloud Hosting" description: "Explainers from the team at Fly.io on how microVMs, agent sandboxes, isolation, and application hosting actually work." --- # Guides to microVMs, Sandboxes, and Cloud Hosting 39 pages, newest first. Each link is the page's Markdown twin; the page itself is the same URL without `.md`, ending in `/`. - [How to Host a Remote MCP Server (and What It Costs)](https://fly.io/learn/host-a-remote-mcp-server.md): Host a remote MCP server over Streamable HTTP: itemized monthly costs vs Cloudflare Workers, wrapping a stdio server, and OAuth end to end. - [Per-PR Preview Environments for Coding Agents](https://fly.io/learn/per-pr-preview-environments-for-coding-agents.md): How to give every pull request, and every coding agent, its own isolated app and database: the GitHub Actions workflow, teardown, cost, and logs. - [Sprites vs E2B vs Modal vs Daytona: Agent Sandboxes Compared](https://fly.io/learn/sprites-vs-e2b-vs-modal-vs-daytona.md): Sprites, E2B, Modal and Daytona compared on isolation, published start times, idle billing, persistence, Docker, GPUs and monthly cost at 1 to 200 sandboxes. - [Taking a Lovable, Bolt or Replit App to Production](https://fly.io/learn/lovable-bolt-replit-to-production.md): How to export a Lovable, Bolt or Replit app, deploy each generated stack, connect your own Supabase project with secrets, and get per-PR previews. - [How to Run LLM-Generated Python Safely](https://fly.io/learn/run-llm-generated-python-safely.md): Run LLM-generated Python in production without exec(): pick an isolation boundary, set network, CPU, timeout and memory limits, and map E2B to Sprites. - [Code Interpreter Pricing: What a Code Execution Sandbox Costs](https://fly.io/learn/code-interpreter-pricing.md): Code interpreter pricing: 1,000 executions priced on OpenAI containers, Bedrock AgentCore, E2B, Daytona, Modal, Vercel, Cloudflare and Fly.io Sprites. - [RL Sandbox Pricing: What Sandboxes for RL Rollouts and Evals Cost](https://fly.io/learn/rl-sandbox-pricing.md): What a sandbox per RL rollout costs: 10,000 agentic rollouts priced on Fly.io Sprites, Prime Intellect, E2B, Daytona, Modal, Vercel and Cloudflare. - [What It Costs to Host an AI Agent 24/7](https://fly.io/learn/always-on-agent-hosting-cost.md): What hosting an AI agent 24/7 costs: a Fly Machine for agents that never sleep, a Sprite for agents that wake on events, and six hosts priced on one workload. - [What It Costs to Run Claude Code or Codex in a Cloud Sandbox](https://fly.io/learn/claude-code-sandbox-cost.md): Claude Code and Codex sandbox cost per session, per developer month and per ticket on E2B, Daytona, Modal, Vercel, Cloudflare and Fly.io Sprites. - [Best Django Hosting Providers in 2026 (Compared)](https://fly.io/learn/django-hosting.md): The best Django hosting providers in 2026 compared: Fly.io, Render, Railway, Heroku, DigitalOcean, PythonAnywhere and Hetzner, by price and upkeep. - [Best Docker Hosting Providers in 2026 (Compared)](https://fly.io/learn/docker-hosting.md): Fly.io, Render, Railway, Cloud Run, Heroku, DigitalOcean, AWS and Hetzner compared for Docker hosting: prices, deploy options and what you still maintain. - [Best Laravel Hosting Providers in 2026 (Compared)](https://fly.io/learn/laravel-hosting.md): The best Laravel hosting in 2026 compared: Fly.io, Laravel Cloud, Forge, Vapor, Render, Railway and a VPS, on queue workers, the scheduler, uploads and cost. - [Best Python Hosting Providers in 2026 (Compared)](https://fly.io/learn/python-hosting.md): The best Python hosting providers in 2026 compared: Fly.io, PythonAnywhere, Render, Railway, Heroku and Cloud Run for Flask, FastAPI, bots and workers. - [Ephemeral vs Persistent Sandboxes: Which One Your Agent Workload Needs](https://fly.io/learn/ephemeral-vs-persistent-sandbox.md): Ephemeral sandboxes reset per task; persistent ones keep files across sessions. How to choose for AI agents, and what burst workloads need. - [Fly.io vs Railway: Deploys, Idle Apps, Databases and Scaling Compared](https://fly.io/learn/fly-vs-railway.md): Fly.io vs Railway for a normal web app: where config lives, how deploys roll out, idle services, managed Postgres, scaling workers, and room to grow. - [Fly.io vs Render: Deploys, Idle Apps, Postgres and Workers Compared](https://fly.io/learn/fly-vs-render.md): Fly.io vs Render for a normal web app: how deploys roll out, what happens when traffic stops, Postgres failover, workers and cron, and room to grow. - [AI Sandbox Pricing, Compared: What Agent Sandboxes Cost and How the Billing Models Differ](https://fly.io/learn/ai-sandbox-pricing.md): AI sandbox pricing compared: E2B, Daytona, Modal, Vercel, Cloudflare and Fly.io Sprites rates, the four billing models, and one agent workload priced end to end. - [Firecracker vs gVisor: Sandbox Isolation for Untrusted and Agent-Generated Code](https://fly.io/learn/firecracker-vs-gvisor.md): Firecracker vs gVisor for running untrusted and agent-generated code: how each isolates, the two threat models, cold start and I/O overhead, and which to pick. - [PaaS Pricing Compared: What Vercel, Render and Heroku Actually Cost](https://fly.io/learn/paas-pricing-compared.md): Vercel, Render, Heroku, Railway and Fly.io pricing side by side: the four billing models, what each plan really includes, and one app priced on all five. - [Where to Run Agent Code: AI Agent Sandbox Providers Compared](https://fly.io/learn/agent-sandbox-providers.md): E2B, Modal, Daytona, Vercel Sandbox, Cloudflare Sandbox and Fly.io Sprites compared on isolation, state between turns, idle behaviour, session caps and egress control. - [Top Heroku Alternatives](https://fly.io/learn/heroku-alternatives.md): Eight Heroku alternatives compared the same way: Render, Railway, DigitalOcean, Vercel, Netlify, Platform.sh, Northflank and Fly.io, with current pricing. - [Top Northflank Alternatives for Secure AI Code Execution and Sandboxed Environments](https://fly.io/learn/top-northflank-alternatives-for-secure-ai-code-execution-and-sandboxed-environments.md): How Fly.io Sprites compare with Northflank for secure AI code execution: isolation, what a sandbox boots from, billing basis, checkpoints, and where each platform is the better choice. - [Fly.io vs Cloudflare: Agent Computers vs Ephemeral Edge Containers](https://fly.io/learn/fly-vs-cloudflare.md): How Fly.io Sprites and Cloudflare's Sandbox SDK differ on disk persistence, sleep behavior, preview URLs, isolation, and where agent state actually lives. - [Fly.io vs Daytona: Lifecycle Timers, Disk Limits, and Isolation Compared](https://fly.io/learn/fly-vs-daytona.md): How Fly.io Sprites and Daytona differ on auto-stop and auto-archive timers, disk size, named isolation, process supervision, and credential handling. - [Fly.io vs E2B: Persistent Agent Computers vs Pause and Resume Sandboxes](https://fly.io/learn/fly-vs-e2b.md): How Fly.io Sprites and E2B differ on pause cost, runtime limits, process supervision, wake behavior, and where agent credentials actually live. - [Fly.io vs Modal: Agent Sandboxes, Snapshots, and Isolation Compared](https://fly.io/learn/fly-vs-modal.md): How Fly.io Sprites and Modal Sandboxes differ on persistence, duration caps, snapshot restore, gVisor vs Firecracker isolation, and GPU workloads. - [Fly.io vs Vercel Sandbox: Persistence, Sessions, and Credentials Compared](https://fly.io/learn/fly-vs-vercel.md): How Fly.io Sprites and Vercel Sandbox differ on process state, session caps, snapshot expiry, wake behavior, and how each keeps credentials out of agent code. - [Best Cloud for Startups: AWS vs GCP vs Azure vs PaaS](https://fly.io/learn/what-is-the-best-cloud-for-startups.md): AWS, Google Cloud, and Azure lead startup hosting, with managed platforms for speed and Fly.io for global compute without hyperscaler overhead - [How Serverless Hosting Scales: Instances, Cold Starts, and Concurrency Limits](https://fly.io/learn/how-does-serverless-hosting-scale.md): Serverless hosting scales by spinning up function instances on demand for each request, then scaling to zero when traffic stops, avoiding idle compute costs - [Node.js Hosting: How to Pick the Right Platform for Your App](https://fly.io/learn/what-is-the-best-nodejs-hosting.md): Node.js hosting depends on your workload: managed PaaS for Git deploys and autoscaling, VPS for root control, serverless for stateless functions - [Postgres Hosting: How to Choose the Right Option for Your App](https://fly.io/learn/what-is-the-best-postgres-hosting.md): The best Postgres hosting depends on workload: managed services handle backups, failover, and scaling, while self-hosted gives full control and more toil - [Cloud Development Environments: What They Fix, and What They Don't](https://fly.io/learn/cloud-development-environment.md): A cloud development environment moves your workspace off your laptop. What CDEs actually fix, where developers say they fall short, and how to decide. - [Golang Hosting: Go Runs as a Single Binary, but Your Host Still Does Real Work](https://fly.io/learn/golang-hosting.md): Golang hosting for statically linked Go binaries: bind ports, supervise processes, manage memory, and ship logs and metrics on Linux or Fly.io - [MicroVM vs Container: How the Kernel Boundary Changes Everything](https://fly.io/learn/microvm-vs-container.md): MicroVM vs container: compare kernel isolation, startup time, memory overhead, and blast radius to choose the right boundary for untrusted code - [Rails Hosting: What Your App Actually Needs to Run in Production](https://fly.io/learn/rails-hosting.md): Rails hosting on Fly.io: learn the runtime, Puma, asset precompilation, background jobs, and deployment details that keep production deploys stable - [Rust Hosting: How to Deploy Compiled Binaries Without the Operational Mess](https://fly.io/learn/rust-hosting.md): Rust hosting on Fly.io: deploy compiled binaries, handle cross-compilation, TLS, restarts, logging, and run Rust backends with Svelte frontends - [What Is a Firecracker microVM?](https://fly.io/learn/firecracker-vm.md): Firecracker VM basics: learn how microVMs use KVM, a reduced device model, and seccomp jailers to run isolated Linux workloads with low overhead - [Agent Sandboxes: Isolated Runtimes for Testing AI Agent Behavior](https://fly.io/learn/agent-sandbox.md): Learn what an agent sandbox is, how a sandbox VM isolates tools and state, and how to test agent behavior without touching production systems - [Virtual Sandbox](https://fly.io/learn/virtual-sandbox.md): What a virtual sandbox is, how it differs from a VM and a container, the isolation ladder from language sandboxes to hardware, and what it does not protect.