$ _
Introduction
TinyComputer is a portable, stateful, container-based operating system designed for Agentic AI. It gives AI agents a persistent, isolated sandbox where they can run code, install dependencies, interact with local files, and communicate with other specialized agent nodes.
Think of it as a tiny computer that your AI agent can call home — a place that remembers everything between sessions, keeps its tools installed, and stays safely isolated from your host system.
The Problem
Today's AI agents suffer from two critical limitations:
- Amnesia. Every session starts from scratch. Installed packages vanish. Files disappear. Context is lost. The agent has to re-do work it already completed.
- Unsafe execution. Running AI-generated code on your host machine is dangerous. Ephemeral cloud sandboxes solve isolation but sacrifice persistence and local access.
- No communication layer. Agents work in isolation. There's no standard way for specialized agents to collaborate, share state, or coordinate tasks.
Visual orchestrators like n8n give you workflow graphs, but their code execution nodes are ephemeral. Cloud sandboxes like E2B or CodeSandbox give you isolation, but state evaporates. TinyComputer bridges the gap.
Features
TinyComputer is being built around these core capabilities:
Persistent State
Every file, every installed package, every configuration change survives across sessions. Your agent picks up exactly where it left off.
Isolated Container Sandbox
Each TinyComputer runs in its own container. AI-generated code executes safely without risking your host system. Full Linux environment with root access inside the sandbox.
Run Code & Install Dependencies
Agents can install any package, compile code, run scripts, and manage services — just like a real computer. Python, Node, Rust, Go, or any language your agent needs.
Local File System Access
Mount host directories into the container. Agents can read your project files, write output, and interact with your local development environment — all within controlled boundaries.
Inter-Agent Communication
TinyComputer nodes can discover and talk to each other. Build agent meshes where specialized agents — a coder, a researcher, a reviewer — collaborate on complex tasks.
Orchestrator Integration
Designed to work alongside visual workflow tools like n8n. Use TinyComputer as the execution backend for your automation pipelines.
How It Works
The architecture is designed around portability and persistence:
┌─────────────────────────────────────────────────────┐ │ YOUR MACHINE │ │ │ │ ┌──────────┐ ┌──────────┐ ┌──────────┐ │ │ │ AI Agent │ │ AI Agent │ │ n8n / │ │ │ │ #1 │ │ #2 │ │Orchestrat│ │ │ └────┬─────┘ └────┬─────┘ └─────┬────┘ │ │ │ │ │ │ │ ▼ ▼ ▼ │ │ ┌────────────────────────────────────────────┐ │ │ │ TinyComputer Runtime │ │ │ │ │ │ │ │ ┌────────────┐ ┌────────────┐ │ │ │ │ │ Container │ │ Container │ ... │ │ │ │ │ (Node A) │◄─► (Node B) │ │ │ │ │ │ │ │ │ │ │ │ │ │ ■ Python │ │ ■ Node.js │ │ │ │ │ │ ■ Files │ │ ■ DB │ │ │ │ │ │ ■ State │ │ ■ State │ │ │ │ │ └──────┬─────┘ └──────┬─────┘ │ │ │ │ │ │ │ │ │ │ ▼ ▼ │ │ │ │ ┌──────────────────────────────────┐ │ │ │ │ │ Persistent Storage Layer │ │ │ │ │ │ (survives across sessions) │ │ │ │ │ └──────────────────────────────────┘ │ │ │ └────────────────────────────────────────────┘ │ │ │ │ ┌──────────────────────────────┐ │ │ │ Host File System (mounted) │ │ │ └──────────────────────────────┘ │ └─────────────────────────────────────────────────────┘
Each container is a fully-functional Linux environment. The runtime manages lifecycle, networking between nodes, and persistent volume mounts. When an agent reconnects, everything is exactly as it was left.
Use Cases
| Who | Use Case |
|---|---|
| AI Developers | Build autonomous agents that maintain state, install tools, and execute code safely across long-running tasks. |
| DevOps Teams | Create reproducible agent environments that can be snapshotted, shared, and deployed consistently. |
| n8n Users | Replace ephemeral code execution nodes with persistent TinyComputer containers that remember previous runs. |
| Researchers | Run long-lived AI experiments where agents accumulate knowledge, data, and installed dependencies over days or weeks. |
| Agent Teams | Spin up multi-agent systems where specialized nodes (coder, tester, reviewer) collaborate through the built-in mesh network. |
Get Notified
TinyComputer is currently under active development. Sign up to be notified when early access is available:
No spam. We'll only email you when there's something worth sharing.