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

Description and use

The author shows how to use agentic-hosting to safely deploy AI‑generated applications on a personal server, giving the agent full control while maintaining isolation and oversight via a REST API.

Problem

Agents can write, test, and fix code but lack a safe, isolated deployment environment, leaving the final deployment step in human hands.

Outcome

The agent can autonomously deploy and manage applications on a dedicated server, reducing human intervention to oversight and security checks.

Workflow

• Install Docker, gVisor, Go, and Nixpacks on a Linux server
• Configure Docker to use gVisor runtime and create a network
• Run a local registry and start the agentic‑hosting service
• Generate a bootstrap token and start the service with it
• Register a tenant via POST /v1/tenants/register to obtain an API key
• Create a service with POST /v1/services
• Trigger a build with POST /v1/services/{serviceID}/builds specifying source_type and source_url
• Provision a database with POST /v1/databases and wire it to the service via POST /v1/services/{serviceID}/env
• Stream build and runtime logs with GET /builds/{buildID}/logs and GET /v1/services/{serviceID}/logs
• Use guardrails such as sandboxing, circuit breakers, and rate limits to ensure safety

Evidence

agentic-hosting is a self-hosted PaaS whose main user is an agent talking to a REST API

Source

One build system (Nixpacks, which turns source code into a container image with no config)

Source

A reconciler checks every 30 seconds that what the database thinks is running matches what Docker is actually running

Source