Th0rgal/sandboxed.sh — Self-hosted orchestrator for AI autonomous agents.

Description and use

**Brief Summary**
Th0rgal / sandboxed.sh is a self‑hosted orchestrator that runs Claude‑code and open‑source AI agents inside isolated Linux workspaces, letting you version‑control skills, configurations, and encrypted secrets with a Git repository. It turns ad‑hoc prompts and tool calls into repeatable, multi‑agent workflows that can be managed via a simple CLI/SDK.

**Value**

• **Isolation & Security** – Each agent runs in its own sandboxed container, preventing cross‑contamination and protecting sensitive data.
• **Version‑controlled Ops** – Skills, prompts, and secrets live in Git, giving you auditability, roll‑backs, and CI/CD‑style deployment of AI logic.
• **Workflow Automation** – By chaining agents and tools, you can build complex pipelines (e.g., data extraction → analysis → report generation) that are reproducible and shareable across teams.

**Practical Adoption Path**

1. **Pilot Setup** – Clone the repo, configure the required Docker/Podman environment, and add your Claude‑code or open‑source model binaries.
2. **Define Skills & Secrets** – Store prompts, tool wrappers, and encrypted secrets in a dedicated Git branch; use the provided CLI to register them.
3. **Create Workflows** – Write YAML/JSON workflow definitions that specify agent order, input/output bindings, and resource limits.
4. **Integrate** – Call the orchestrator from existing CI pipelines, internal APIs, or a custom SDK to trigger workflows on demand.
5. **Scale** – Deploy the orchestrator on a Kubernetes node pool or a VM fleet; the built‑in health checks and metrics let you monitor performance and add more sandbox instances as load grows.

**Production Readiness**

• **Activity & Community** – 422 stars, 44 forks, recent commits (as of 2026‑05‑11), and a Rust codebase indicate active maintenance and a growing user base.
• **Maturity** – The project already exposes a stable CLI/SDK, supports encrypted secret handling, and follows a Git‑Ops model, which aligns with typical enterprise DevOps practices.
• **Risks to Verify** – Final due‑diligence should confirm the license compatibility, review any open security issues, and ensure at least one maintainer is responsive for critical patches.

Overall, sandboxed.sh is production‑grade for a serious pilot, especially for teams that need secure, reproducible AI agent pipelines and want to manage them through familiar Git‑centric workflows.

Problem

Th0rgal/sandboxed.sh helps turn isolated prompts and tools into repeatable agent workflows.

Outcome

High for an OSS candidate: recent activity, adoption, and ecosystem signals are strong enough for a serious pilot.

Workflow

• coordinate multi-agent workflows
• add tool-use pipelines
• standardize agent memory

Evidence

Self-hosted orchestrator for AI autonomous agents. Run Claude Code & Open Code in isolated linux workspaces. Manage your skills, configs and encrypted secrets with a git repo.

Source