An interesting new project has arrived that lets you connect your developer assistant directly to your server. It's about OpenShip (oblien/openship), a recently released self-hosted PaaS. The most exciting aspect of this project is that it supports the Model Context Protocol (MCP) by default, which is a major hot topic these days.

Previously, giving an agent (like Claude Code or Cursor) control over your infrastructure meant handing over complex SSH keys or opening up risky permissions. However, OpenShip uses a local CLI as a bridge to communicate with the agent. By leveraging the session token the developer has already obtained via their local terminal login, you can safely delegate tasks like deployment, database provisioning, and rollbacks to the agent.

Just a heads-up, there's a commerce-focused OpenShip (openshiporg) project on GitHub with a similar name, so be careful not to mix them up. For the infrastructure-related OpenShip, you can simply register your CLI commands in the configuration file as shown below, and the agent will be able to recognize them immediately.

{
  "mcpServers": {
    "openship": {
      "command": "openship",
      "args": ["mcp", "start"]
    }
  }
}

Once integrated, you can give the agent natural language commands like "Check the server build logs for me" or "Roll back to the previous build and restart the DB." It feels strongly that development tools are evolving into a form that is "easier for agents to use" rather than just humans. I'm really looking forward to seeing how the server and backend infrastructure space is reshaped around agent-native interfaces.