sandboxed.sh  by Th0rgal

Self-hosted orchestrator for AI autonomous agents

Created 2 months ago
252 stars

Top 99.6% on SourcePulse

GitHubView on GitHub
Project Summary

Summary

sandboxed.sh is a self-hosted orchestrator for AI autonomous agents, enabling diverse coding agents (Claude Code, OpenCode, Amp) to run in isolated Linux workspaces. It simplifies managing agent skills, configurations, and secrets via a Git repository, facilitating unattended, secure, and version-controlled AI operations for automating complex tasks or processing sensitive data locally.

How It Works

The system orchestrates AI agents within isolated Linux environments using systemd-nspawn (native) or Docker containers. It supports multiple agent runtimes and leverages a Git-backed "Library" for version-controlled configurations, skills, and tools. This approach ensures robust workspace isolation and reproducible deployments across diverse AI agent ecosystems.

Quick Start & Requirements

Recommended installation is Docker (~5 min setup) for any OS. Native installation targets Ubuntu 24.04 LTS (~30 min setup).

  • Docker: git clone, cd, cp .env.example .env, docker compose up -d. Access http://localhost:3000. Note: privileged: true required for container workspace isolation.
  • Native: Requires Ubuntu 24.04 LTS; involves git pull and cargo build or dashboard updates.
  • Links: Website, Discord, Getting Started, Docker Guide, Native Guide.

Highlighted Details

  • Multi-Runtime Support: Integrates Claude Code, OpenCode, and Amp agents.
  • Mission Control: Remote agent management with real-time streaming.
  • Isolated Workspaces: Secure, containerized Linux environments.
  • Git-backed Library: Centralizes and version-controls agent configurations.
  • MCP Registry: Supports optional external tool servers (e.g., desktop automation).
  • Multi-platform UIs: Next.js web dashboard and SwiftUI iOS app.
  • OpenAI-compatible Proxy: Standard API call compatibility.

Maintenance & Community

Marked "Work in Progress" and under active development, welcoming contributions. Community support via Discord.

Licensing & Compatibility

Permissive MIT license, suitable for commercial and closed-source applications.

Limitations & Caveats

Active development implies potential instability or breaking changes. Docker requires privileged: true for container workspaces (security implications). Native installation is Ubuntu 24.04 LTS specific.

Health Check
Last Commit

1 day ago

Responsiveness

Inactive

Pull Requests (30d)
206
Issues (30d)
77
Star History
139 stars in the last 30 days

Explore Similar Projects

Starred by Lilian Weng Lilian Weng(Cofounder of Thinking Machines Lab), Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), and
59 more.

AutoGPT by Significant-Gravitas

0.1%
182k
AI agent platform for building, deploying, and running autonomous workflows
Created 2 years ago
Updated 17 hours ago
Feedback? Help us improve.