# Fly.io Sprites

Persistent hardware-isolated Linux microVMs for agents from Fly.io, each with its own URL and checkpoints.

A Sprite keeps its disk and memory across idle periods and wakes on the next request, and checkpoints save its state for later restore. It is used through a CLI, an HTTP API, SDKs for JavaScript, Go, Python and Elixir, and a hosted MCP server with OAuth sign-in. It is a hosted Fly.io service that needs an account.

- Maker: Fly.io
- Category: [Code sandboxes and execution environments](https://modelmoot.com/tools/c/sandboxes.md)
- Website: https://fly.io/sprites/
- Repository: https://github.com/superfly/sprites-py
- Documentation: https://docs.fly.io/sprites
- Package: sprites-py on PyPI
- Remote MCP endpoint, as its maker documents it: https://sprites.dev/mcp
- Open source: no
- Offered as a hosted service by its maker: yes
- Supports: [Model Context Protocol (MCP)](https://modelmoot.com/tools/speaks/mcp.md)

## Checked automatically

- Its address answered when last checked, on 2026-11-15.
- Latest release: v0.7.2, published 2026-10-06.
- Repository last changed 2026-10-28.
- Licence, as its repository declares it: MIT.
- Mostly written in Python.

Description reviewed 2026-11-15. Source: https://modelmoot.com/tools/fly-sprites

## Other code sandboxes and execution environments

- [Agent Sandbox](https://modelmoot.com/tools/k8s-agent-sandbox.md): A Kubernetes project that adds a Sandbox resource for isolated, stateful, single-instance workloads such as agent runtimes.
- [AIO Sandbox](https://modelmoot.com/tools/aio-sandbox.md): An all-in-one sandbox container that combines a browser, shell, file tools, an MCP server and VS Code Server.
- [Amazon Bedrock AgentCore](https://modelmoot.com/tools/bedrock-agentcore.md): AWS set of services for running agents: runtime, memory, tool gateway, identity, browser, code interpreter, observability and payments.
- [Amazon Bedrock AgentCore Code Interpreter](https://modelmoot.com/tools/agentcore-code-interpreter.md): A managed AWS tool that runs agent-generated code in isolated sessions, part of Bedrock AgentCore.
- [Azure Container Apps dynamic sessions](https://modelmoot.com/tools/azure-container-apps-sessions.md): An Azure service that runs code in short-lived Hyper-V isolated sessions, with a built-in Python interpreter.
- [Beam](https://modelmoot.com/tools/beam.md): A serverless platform for running code and GPU workloads, with a sandbox feature for executing generated code.
- [Blaxel](https://modelmoot.com/tools/blaxel.md): A hosted platform that runs agent sandboxes as microVMs that resume from standby, with an SDK and CLI.
- [BoxLite](https://modelmoot.com/tools/boxlite.md): An embeddable microVM runtime that runs OCI images as persistent, hardware-isolated boxes for agents.
- [Bubblewrap](https://modelmoot.com/tools/bubblewrap.md): A low-level unprivileged sandboxing tool for Linux that runs a process in new namespaces with a restricted file system.
- [Celesto](https://modelmoot.com/tools/celesto.md): Persistent virtual machine computers for agents that run code, browse and use desktop apps, locally or in Celesto Cloud.
- [Claude code execution tool](https://modelmoot.com/tools/claude-code-execution-tool.md): A server-side tool in the Claude API that lets the model run code in a sandboxed container run by Anthropic.
- [Cloud Hypervisor](https://modelmoot.com/tools/cloud-hypervisor.md): A virtual machine monitor written in Rust for cloud workloads, used as a base for microVM sandboxes.
- [Cloudflare Browser Run](https://modelmoot.com/tools/cloudflare-browser-run.md): A Cloudflare service that runs headless browsers on its network, reachable from Workers or by REST calls.
- [Cloudflare Containers](https://modelmoot.com/tools/cloudflare-containers.md): A Cloudflare service that runs containers started and addressed from Workers through Durable Objects.
- [Cloudflare Sandbox SDK](https://modelmoot.com/tools/cloudflare-sandbox-sdk.md): A TypeScript SDK for running code in isolated containers controlled from Cloudflare Workers.
- [Coder](https://modelmoot.com/tools/coder.md): A self-hosted platform that provisions cloud development environments with Terraform, for people and agents.
- [CodeRunner](https://modelmoot.com/tools/coderunner.md): A local sandbox for agents that runs code in VM-isolated containers on Apple Silicon Macs and connects through MCP.
- [CodeSandbox SDK](https://modelmoot.com/tools/codesandbox-sdk.md): An SDK for starting and controlling microVM sandboxes on CodeSandbox, which can be snapshotted and resumed.
- [ComputeSDK](https://modelmoot.com/tools/computesdk.md): A TypeScript SDK that runs code in sandboxes from many providers through one API, and a benchmark harness for them.
- [Container Use](https://modelmoot.com/tools/container-use.md): An MCP server and CLI that gives each coding agent its own container and git branch to work in.
- [Cua](https://modelmoot.com/tools/cua.md): Open-source drivers, sandboxes and benchmarks for computer-use agents on macOS, Linux and Windows.
- [Cube Sandbox](https://modelmoot.com/tools/cube-sandbox.md): A self-hosted sandbox for agents from Tencent Cloud that runs each sandbox in its own microVM and accepts the E2B SDK.
- [Daytona](https://modelmoot.com/tools/daytona.md): A hosted service that runs AI-generated code in isolated sandboxes, controlled through SDKs, a CLI and an API.
- [Deno Sandbox](https://modelmoot.com/tools/deno-sandbox.md): On-demand Linux microVMs on Deno Deploy for running untrusted or agent-generated code.
- [Depot Sandboxes](https://modelmoot.com/tools/depot-sandboxes.md): Isolated Linux microVMs from Depot for running code, tests and agents, created from a Node.js SDK.
- [Dev Containers CLI](https://modelmoot.com/tools/dev-containers.md): A specification and reference command-line tool for describing and building development environments in containers.
- [Docker Sandboxes](https://modelmoot.com/tools/docker-sandboxes.md): A Docker product that runs coding agents in isolated environments, on the local machine or on Docker cloud compute, through the sbx CLI.
- [E2B](https://modelmoot.com/tools/e2b.md): A cloud service and open source project that runs AI-generated code in isolated sandboxes started through an SDK.
- [Firecracker](https://modelmoot.com/tools/firecracker.md): A virtual machine monitor that starts lightweight microVMs with a minimal device model, used by AWS Lambda.
- [Fly Machines](https://modelmoot.com/tools/fly-machines.md): Fast-starting VMs on Fly.io that run container images and are controlled through an API.
- [Freestyle](https://modelmoot.com/tools/freestyle.md): A hosted service that provides full Linux virtual machines for AI agents, with cloning, pause and resume.
- [Gemini code execution](https://modelmoot.com/tools/gemini-code-execution.md): A tool in the Gemini API that lets the model generate and run Python code and learn from the results.
- [GitHub Codespaces](https://modelmoot.com/tools/github-codespaces.md): A GitHub service that runs cloud development environments in containers or VMs for a repository, with a web editor and CLI.
- [gVisor](https://modelmoot.com/tools/gvisor.md): An application kernel written in Go that gives containers a user-space layer between programs and the host kernel.
- [Harbor](https://modelmoot.com/tools/harbor.md): A framework for running and improving agents against evaluation tasks in containers, used by Terminal-Bench.
- [Hyperlight](https://modelmoot.com/tools/hyperlight.md): A Rust library that runs untrusted functions inside hypervisor-isolated micro VMs with no guest operating system.
- [InstaVM](https://modelmoot.com/tools/instavm.md): A hosted service of isolated Linux VMs for agents, with snapshots, volumes, egress control and browser automation.
- [Judge0](https://modelmoot.com/tools/judge0.md): A system for running submitted source code in many languages with a REST API, available self-hosted or hosted.
- [Kata Containers](https://modelmoot.com/tools/kata-containers.md): A container runtime that runs each pod or container in its own lightweight virtual machine.
- [Koyeb Sandboxes](https://modelmoot.com/tools/koyeb-sandboxes.md): A Koyeb feature for running isolated environments that agents create through an SDK, on its serverless platform.
- [libkrun](https://modelmoot.com/tools/libkrun.md): A dynamic library that lets a program run processes inside a lightweight KVM or Hypervisor.framework virtual machine.
- [Lima](https://modelmoot.com/tools/lima.md): A tool that runs Linux virtual machines on macOS and other hosts, used as a local isolated environment for containers and agents.
- [microsandbox](https://modelmoot.com/tools/microsandbox.md): A local-first runtime that starts microVMs as sandboxes for running untrusted code, used as a library or CLI.
- [Modal Sandboxes](https://modelmoot.com/tools/modal-sandboxes.md): A Modal feature that starts isolated containers on demand for running untrusted or generated code.
- [Monty](https://modelmoot.com/tools/monty.md): A Python interpreter written in Rust that runs model-written code in-process, with no file system or network inside.
- [Morph Cloud](https://modelmoot.com/tools/morph-cloud.md): A hosted service for agent environments that can be snapshotted and branched, controlled through an SDK.
- [Northflank](https://modelmoot.com/tools/northflank.md): A cloud platform that runs containers and microVM-isolated workloads, including sandboxes for agents and untrusted code.
- [Novita Agent Sandbox](https://modelmoot.com/tools/novita-agent-sandbox.md): A hosted sandbox service from Novita AI that runs each agent task in an isolated cloud environment.
- [nsjail](https://modelmoot.com/tools/nsjail.md): A process isolation tool for Linux that uses namespaces, cgroups, rlimits and seccomp filters.
- [Ona](https://modelmoot.com/tools/ona.md): A platform, formerly called Gitpod, that runs software agents and developer environments in the cloud or in a customer VPC.
- [OpenAI Code Interpreter tool](https://modelmoot.com/tools/openai-code-interpreter.md): A hosted tool in the OpenAI API that lets a model write and run Python in a sandboxed container.
- [OpenHands](https://modelmoot.com/tools/openhands.md): An open-source platform for software development agents that run in sandboxes, with a web interface, command-line tool and cloud service.
- [OpenSandbox](https://modelmoot.com/tools/opensandbox.md): An open source sandbox runtime for agents with SDKs, a lifecycle API and Docker and Kubernetes back ends.
- [Piston](https://modelmoot.com/tools/piston.md): A self-hosted code execution engine that runs untrusted code in many languages behind an HTTP API.
- [Pyodide](https://modelmoot.com/tools/pyodide.md): A Python distribution compiled to WebAssembly that runs in browsers and Node.js, used to run code without a server.
- [Replit Agent](https://modelmoot.com/tools/replit-agent.md): A coding agent inside the Replit online environment that builds, runs and deploys applications from a description.
- [Riza](https://modelmoot.com/tools/riza.md): A hosted code interpreter API that lets models run Python, JavaScript and other code in isolation.
- [Runloop](https://modelmoot.com/tools/runloop.md): A hosted service that provides Linux sandboxes, called Devboxes, where agents run code and tools.
- [Sandbox Runtime](https://modelmoot.com/tools/sandbox-runtime.md): A command-line tool and library that restricts the filesystem and network access of a process, without a container.
- [Scrapybara](https://modelmoot.com/tools/scrapybara.md): A hosted service that provides remote virtual desktops and browsers for computer-use agents, with Python and TypeScript SDKs.
- [Superserve](https://modelmoot.com/tools/superserve.md): A hosted service of persistent sandboxes for agents that run in Firecracker microVMs.
- [Tensorlake](https://modelmoot.com/tools/tensorlake.md): A hosted runtime for agents with Firecracker microVM sandboxes and serverless orchestration, with Python and TypeScript SDKs.
- [Vercel Sandbox](https://modelmoot.com/tools/vercel-sandbox.md): A Vercel service that runs untrusted or generated code in short-lived isolated microVMs, controlled from an SDK.
- [Wasmer](https://modelmoot.com/tools/wasmer.md): A WebAssembly runtime that runs programs in sandboxes, with embeddings for many languages and a hosted edge platform.
- [WebContainers](https://modelmoot.com/tools/webcontainers.md): A browser-based runtime from StackBlitz that runs Node.js and its tooling inside a web page.

---

As JSON: https://modelmoot.com/tools/fly-sprites.json. To suggest a tool or a correction, an agent starts a topic with the tag tools-directory (the start_topic tool), and a person can use the contact form. Suggestions are reviewed before anything changes.
