Agents
RailPush agents are packaged AI assistants that run as managed RailPush services. They use the same plans, logs, health checks, billing controls, audit history, and workspace permissions as any other service.
Use this page when you want to launch a ready-to-use agent product. Use the MCP Server page when you want to connect an external client such as Claude Code, Cursor, or your own agent runtime.
Products
| Product | Best for | Authentication | Launch |
|---|---|---|---|
| OpenClaw Cloud | Personal assistant with browser chat, Telegram, or Discord | Provider API key or interactive ChatGPT OAuth | Guided dashboard wizard |
| Hermes Agent | Developer agent with GPT OAuth and an OpenAI-compatible API | ChatGPT OAuth by default | Guided dashboard wizard |
OpenClaw Cloud
OpenClaw is the marketable personal AI operator path. It creates a hosted assistant with browser chat, optional Telegram or Discord, persistent workspace files, and optional RailPush workspace access.
Choose OpenClaw when the user wants an assistant that can chat in a browser or messaging channel and can optionally operate services, databases, logs, and deploys through a scoped RailPush key.
Hermes Agent
Hermes is the frictionless GPT OAuth developer-agent path.
It creates a managed Hermes Agent runtime with a RailPush guest chat URL and an OpenAI-compatible /v1/chat/completions endpoint.
Choose Hermes when the user wants GPT OAuth by default and a hosted Nous Hermes Agent without configuring an OpenAI API key.
Billing and safety
- Agent products only launch on paid fixed monthly plans.
- Workspace access uses scoped RailPush credentials and can be revoked.
- Agent-created paid resources should ask the user for confirmation before launch.
Next steps
- Launch OpenClaw Cloud
- Launch Hermes Agent
- Connect an external client with MCP Server