AI Agents
AI Agents
The AI agents on RepoCloud and how to choose between them: the DevOps AI Agent, agent apps, and your own agent
AI Agents on RepoCloud
“AI agent” means three different things on RepoCloud, and they share a name and little else. The DevOps AI Agent is RepoCloud’s own: an AI system administrator that works on your server over SSH while you chat with it. Agent apps are AI agents, such as OpenClaw and Hermes Agent, that run on your own server, and the DevOps AI Agent can install them for you. Your own agent, in Cursor, Claude Code, or another tool, can create and manage RepoCloud servers through the API and MCP.
An AI system administrator that installs, configures, and changes an application on your own VPS while you chat with it
AI agents that you run on your own VPS, such as OpenClaw and Hermes Agent
Let Cursor, Claude Code, or another AI agent create and manage RepoCloud servers through the API and MCP
Which One Do I Need?
| I want to… | Use | Read |
|---|---|---|
| Put a public GitHub project on a server and have it set up for me | The DevOps AI Agent, through AI Deploy | AI Deploy |
| Run a catalog app on its own VPS, with root access, and have it set up for me | The DevOps AI Agent, through the Deploy as VPS with AI system admin agent switch | Deploy as VPS |
| Keep changing an app that the agent set up: settings, plugins, upgrades | The DevOps AI Agent, through Resume Chat on the server’s manage page | Customizing Applications |
| Run an AI assistant of my own, with a web interface or chat apps | An agent app: OpenClaw or Hermes Agent, installed through AI Deploy | Agent Apps |
| Let my own AI tools create and manage servers | The AI Agent Portal, with the API or MCP | Your Own Agent |
| Deploy from a private repository | Your own agent with MCP, because AI Deploy takes public repositories | Your Own Agent |
| Run a catalog app in one click, with no AI involved | A Container App | Deployment Options |
How They Compare
| The DevOps AI Agent | Agent apps | Your own agent | |
|---|---|---|---|
| What it is | RepoCloud’s AI system administrator for your server | OpenClaw and Hermes Agent, open-source agents that run on a VPS App | Any AI agent that can use MCP with a custom header, or call the REST API |
| Where you start | AI Deploy, the Deploy as VPS switch, or Resume Chat on the manage page | AI Deploy, with the app’s repository | The AI Agent Portal in your dashboard, where you create an API key |
| Who does the work | The agent, over SSH, on your server | The agent software you deployed, on the tasks you give it | Your agent, calling the RepoCloud API, and SSH if it needs it |
| Which AI model | RepoCloud chooses it | The provider whose key you give it, such as Anthropic or OpenRouter | The one your tool uses |
| What you pay RepoCloud | The server, by the hour; the agent’s AI usage is not billed separately | The server, by the hour; model use is billed by your model provider | The servers it creates, by the hour |
| Servers it works on | A server created through AI Deploy | The server the app runs on | Ubuntu VPS servers created through the API |
Before You Start
- Every route needs a balance above $3. AI Deploy (which is also how OpenClaw and Hermes Agent are installed), the Ubuntu VPS listing, and a deploy through the API all create a VPS. A new API key also needs a payment of $1 or more on the account.
- An AI agent can be wrong. It is a language model working from documentation, and it can misjudge a step. Check what it reports, and on a VPS take a snapshot before a risky change. Safety and Control explains how.
- What you type reaches an AI model. Whatever you write to an agent, and whatever it reads on your server, is processed by an AI model from a provider. Give an agent the secrets its job needs and no others.
Next Steps
What the agent does on your server and how to work with it
Recipes for changing an application with the agent
Snapshots, Stop, what the agent can reach, and who sees what
Paste a GitHub URL and start a deployment