I Gave My Coding Agent a Server of Its Own. Here’s How to Do It Safely.

Description and use

Using the self‑hosted PaaS agentic‑hosting, developers can safely deploy and manage code generated by AI agents with isolated containers and strict guardrails.

Problem

Agents can write and deploy code but lack safe, isolated infrastructure to run on, forcing developers to manually manage deployment and security.

Outcome

Agentic‑hosting enables coding agents to autonomously build, deploy, and manage applications on a dedicated server with strong isolation and guardrails, reducing manual intervention and improving security.

Workflow

• Install Docker, gVisor, Go, and Nixpacks on a Linux server.
• Configure Docker to use gVisor runtime and set address pools.
• Create Traefik network and run local registry.
• Build the agentic‑hosting binary and generate master key and bootstrap token.
• Run the service as systemd and verify health endpoint.
• Register a tenant via API to obtain API key.
• Create services and trigger Nixpacks builds from git repositories.
• Provision databases and set environment variables via API.
• Deploy services and access them through Traefik‑managed subdomains.
• Use Claude Code skill or bash scripts to automate deployment flows.
• Monitor guardrails such as reconciler, circuit breaker, disk watermarks, rate limits, idempotency, and backups.

Evidence

agentic-hosting is an attempt at a third option. It's a self-hosted PaaS for your own bare-metal box where the primary user is an AI agent and the primary interface is a REST API.

Source

Secure by default. Every tenant container runs under gVisor, with a read-only root filesystem, all Linux capabilities dropped, and no-new-privileges. Secrets are encrypted with AES-256-GCM.

Source

A circuit breaker stops restarting a service after 5 crashes in 10 minutes, so a confused agent can't thrash your box.

Source