Profile
Back to NewsBack
GitHub Trending 10 min
Reader Mode
tastyeffectco/sandboxd: Open-source, self-hosted AI app builder — an agent builds real apps in isolated sandboxes on your own server, each live at a preview URL. Self-host in one command. MIT.

tastyeffectco/sandboxd: Open-source, self-hosted AI app builder — an agent builds real apps in isolated sandboxes on your own server, each live at a preview URL. Self-host in one command. MIT.

10 hours ago

sandboxd

Your own AI agents, working for you around the clock, on your computer.
They take on jobs, keep at them, report back, and build the apps you need along the way.
Free to start · No code · Runs on your machine · Open source (MIT)

🔥 Trend of the Day on Trendshift

sandboxd • Trend of the Day on Trendshift

Live demo   Get started   Star sandboxd

Docs License: MIT Runs on Docker Release Sponsor

sandboxd: the Hermes agent watching flight prices and planning meals, in front of an app it uses


What is sandboxd?

An AI crew that works for you: agents first, apps on demand.

Chatbots answer and forget. sandboxd's agents keep working: they remember you, run jobs on a schedule, browse the web, message you on Telegram, and use (or build) apps to get things done. You talk to them in plain words.

| You say… | Your crew… | |---|---| | "Watch flight prices to Marrakech and tell me when it drops under €120" | an agent checks every morning and messages you the moment it does | | "Plan my dinners every week from my recipes" | an agent makes the plan every Sunday and sends it to you | | "Build me a recipe app with a shopping list" | a coding agent builds a real, working app in minutes, with its own link | | "Install Excalidraw for me" | sets up a popular free app in one click |

Everything runs on your computer or server: your apps, your code and your data live there, and no subscription is needed. It starts free with a built-in AI, and you can plug in ChatGPT, Claude, any other AI, or a model running on your own computer whenever you like.

How it works

flowchart LR
    you(["🧑 You"]) -- "plain words" --> agent["🤖 sandboxd<br/>(Home)"]
    agent -- "hand off a job" --> crew["🛰️ Your agents<br/>Hermes · OpenClaw · OpenMausBot"]
    crew -- "report back" --> tg["📲 You, on Telegram<br/>or in the chat"]
    agent -- "build an app" --> sb["📦 Private sandbox<br/>a coding agent writes it"] --> url["🌐 Your app<br/>at its own link"]
    crew -. "use and update" .-> url
    agent -- "install a free app" --> free["🧩 Open-source app<br/>ready to use"]
  1. Ask on Home. One chat with the sandboxd agent, your crew's lead. It
hands each job to the right agent, or starts building.
  1. Your agents get to work, and keep at it. Each runs in its own locked-down
container with its own memory, on a schedule or whenever you ask, and tells you what it found.
  1. Apps when you need them. A coding agent builds each app in a private
sandbox while you watch the live preview. Chat to change anything; every step is saved, so you can go back.

An app built from one sentence, live next to its chat

A quick tour

Home: one chat with the sandboxd agent
Home
One chat with your crew's lead: hand off a job, ask for an app, or ask what your agents did today.
Add an agent
Your agents
Hermes, OpenClaw and OpenMausBot, or several of each, each with its own memory.
A chat with Hermes
Agents that keep working
Hermes watches flight prices every morning, then plans dinners and fills the shopping list in your Recipe Box app.
My apps
My apps
Every app your crew built, plus what's trending on GitHub.
Free apps
Free apps
Excalidraw, draw.io, Uptime Kuma and many more, installed in one click.
Settings: choose your AI
Choose your AI
Free to start. Add ChatGPT, Claude, OpenRouter, CometAPI or a model on your own computer.

▶ Try the live demo · click around with sample data, nothing to install
Light or dark: Settings → Appearance follows your computer, or pick one.

What your crew can do

  • 🤖 Agents that work around the clock. Open-source assistants
(Hermes, OpenClaw, OpenMausBot) that remember you, run scheduled jobs, browse the web, reach you on Telegram, and create or change apps to get the job done. Add several of each. Only agents that run fully on your machine, each in its own locked-down container with no keys inside. Agents →
  • 💬 One lead for the whole crew. The sandboxd agent hands jobs to your
agents, brings the answers back, and never deletes anything without your yes. The sandboxd agent →
  • ✨ Apps built on demand. A habit tracker, a landing page for your bakery, a
team dashboard. A coding agent builds it in minutes; attach your logo or photos and it uses them.
  • 🧩 Free apps in one click. Popular open-source apps (whiteboards,
diagrams, notes, PDF tools, a CAD modeler, a physics simulator…), each tested and ready. Free apps →
  • 🧠 Pick your AI. Free by default. In Settings → AI, connect your
ChatGPT or Claude subscription, OpenRouter, CometAPI, or a model on your own computer (Ollama, LM Studio). Model connections →
  • 🚀 Keep what you build. The code is yours: push it to your own Git
repository or run it on your own server.

Get started

🖥️ Windows or macOS: the desktop app. Click Install & start and it sets everything up by itself: no Docker, no terminal, no account. Download and setup →

🐧 Linux, a server, or your own machine: one command. Needs Docker (with the Compose plugin) and git.

curl -fsSL https://raw.githubusercontent.com/tastyeffectco/sandboxd/main/install.sh | bash

It prints your link: open it, create a password, and tell sandboxd what you want on Home: hand off a job or ask for an app. Works on Intel/AMD and ARM, including Apple Silicon.

☁️ No computer to spare? Deploy to a VPS in one click or let us run it for you.

For developers

sandboxd is also an engine you can build on. Everything the console does is a /v1 HTTP call, so you can run many sandboxes for your own users: an AI app-builder, an agent platform, a coding playground, or per-user preview environments.

API=http://127.0.0.1:9090
curl -s -XPOST $API/v1/agents/claude-code/api-key -d '{"api_key":"sk-ant-..."}'
ID=$(curl -s -XPOST $API/sandbox -d '{"ports":[3000]}' | sed -E 's/."id":"([^"]+)"./\1/')
curl -s -XPOST $API/v1/sandboxes/$ID/tasks -d '{"prompt":"build a todo app on port 3000","agent":"opencode"}'

open the result at http://s-$ID-3000.preview.localhost

Under the hood

  • One Go program driving Docker, with Traefik for links and SQLite for state:
no Kubernetes, no separate database, no queue.
  • Isolated sandboxes: a hardened container per app, with its own files,
limits and live preview link. Idle apps sleep and wake on demand, so one ordinary machine holds many.
  • Coding agents: OpenCode, Claude Code, Codex and Pi. **No credential ever
enters a sandbox** (a proxy adds it on the wire), and every task is checkpointed and revertible.
  • Bring any model: OpenAI- and Anthropic-compatible endpoints, local servers
and LiteLLM gateways (including AWS Bedrock). Keys stay encrypted on the control plane.
  • Runtime presets: React/Vite, Next.js, Node/Express, FastAPI, Worker.
Base image, stacks & presets →
  • Files, Git and secrets: in-browser editor and diffs, commit and push,
encrypted write-only secrets per app. Snapshots, fork and restore, activity timeline, per-process logs. Architecture →

Install options (verified images, build from source, headless, upgrades)

  • Console: http://console.localhost. Run ./console-login.sh to show
the URL or reset the password.
  • API: http://127.0.0.1:9090 (curl http://127.0.0.1:9090/healthz → ok)
  • Verified images: on a clean stable release checkout with cosign
installed, the installer pulls the prebuilt multi-arch images and verifies each one is cosign-signed with build provenance by the release workflow at the exact release tag. ./install.sh --pull requires those checks; a failure in automatic mode falls back to building from source.
  • Build from source: ./install.sh --build compiles every image locally,
so you audit exactly what runs. A first-class, fully supported path.
  • Desktop app from the CLI: installed by default on macOS; opt out with
--no-desktop / -NoDesktop or SANDBOXD_DESKTOP=0. install.ps1 remains for existing Docker Desktop/WSL setups on Windows.
  • Headless: --no-console --no-desktop (Windows: -NoConsole -NoDesktop).
  • Upgrade later: ./upgrade.sh backs up your database, health-checks the
new version and rolls back automatically if it fails (Upgrading). ./upgrade.sh --check shows your version.

Full walkthrough → sandboxd.io/quickstart · Console guide · AGENTS.md for driving it from your own agent. Just need one or two containers, with no AI? A shell script or docker run is simpler (why not just a script?).

🚀 Deploy to a VPS in one click

sandboxd needs one Linux server with Docker — nothing else. Grab a server below (2 vCPU / 4 GB is plenty to start), paste our cloud-init file at creation, and it installs itself.

Disclosure: we earn a referral commission when you sign up through the
provider links below, at no additional cost to you. It helps fund sandboxd's
development.

Deploy on Vultr</a>

Then on the new server (or paste deploy/cloud-init.yaml as user-data and skip this):

curl -fsSL https://raw.githubusercontent.com/tastyeffectco/sandboxd/main/deploy/bootstrap.sh | sudo bash

Full per-provider walkthrough: deploy/DEPLOY.md.

Get a public URL — sandboxd Cloud

Out of the box, previews live on .sslip.io or localhost: fine on your network, useless for sharing. sandboxd Cloud gives the server you already run https://you.sandboxd.io (console, API and *.preview.you.sandboxd.io) in one command — no domain, DNS, open ports or certificates, and it works behind NAT, on a laptop, or on a $5 VPS. Your compute and data stay on your machine. $9.99/server/month, 50% off at launch. Prefer your own domain? Production / TLS.

Managed — we run it for you

Don't want to run the server yourself? We'll **install, configure, and manage sandboxd on your own box** — BYOC (your compute, your data, no lock-in). One-time setup + optional monthly management. See Managed →


📣 Stay in the loop

Self-hosting is MIT and free, forever — git clone and you're done.
Want the 1.0 release + occasional product updates, or a managed cloud so you don't run the box yourself? Two links:

Get 1.0 + release news   sandboxd Cloud waitlist

No spam, one-click unsubscribe, and the cloud is optional — the self-hosted engine is always the full product.


Documentation

Full docs live at sandboxd.io:

| Getting started | Guides | Reference | |---|---|---| | What is sandboxd? | The web console · Get an app running | API (OpenAPI) | | Quickstart | Coding agents | Configuration | | Core concepts | Base image, stacks & presets | Architecture | | Roadmap | Presets & sandbox.yaml · Auto-detection & App Store · Undo a task · Production / TLS · Hardening · Uninstall | AGENTS.md · ARCHITECTURE.md |

In this repository: desktop · free apps · agents · the sandboxd agent · model connections · agent auth.

Building on it from your own agent? AGENTS.md is a copy-pasteable runbook.

Roadmap & community

Beta · 0.x. Container isolation (not VMs), single-server, and API auth off
by default — all fine for your own users; tighten before untrusted
multi-tenancy. See Hardening. Expect the
occasional breaking change before 1.0 — pin a version and update as you go.

⭐ If sandboxd is useful, star it

Stars are how other builders find sandboxd — it's the fastest way to support the project (and it keeps us going). Thank you 🙏

Star History Chart</a>

License

MIT. Use it, ship it, sell what you build on it.

Sponsors

sandboxd is free and MIT-licensed. **Sponsors keep it maintained and fund the deploy roadmap** — thank you to everyone who chips in.

No sponsors yet — be the first.

Chat with me