Introduction

cloudrun is a control plane for deploying your company's AI brains and managing the fleet of agents that run them — from one place.

Each project gets its own isolated machine running a real agent runtime. cloudrun provisions it, configures it, and gives you one UI over the whole fleet instead of a dashboard per agent.

What cloudrun does

• Deploy brains — stand up a project on managed infrastructure, no DevOps.
• Manage the fleet — every agent across every project on one canvas.
• Connect tools — hundreds of third-party services through one-click connectors.
• Automate — scheduled tasks and messaging channels per agent.
• Control access — organizations, members, roles, and API keys.

How it fits together

An organization owns projects. A project runs one or more engines — agent runtimes, each kind on its own substrate — and each engine hosts agents, each with its own instructions, skills, tools, channels, and schedules.

organization
└── project
    └── engine  (an agent runtime — Hermes today, more kinds to come)
        └── agent  (instructions, skills, tools, channels, automations)

Open source

cloudrun is MIT licensed and self-hostable — see Self-hosting to run it on your own infrastructure. Source: github.com/shawnxman1423/openbrain.

Next steps

• Quickstart — create your workspace and your first agent
• Concepts — projects, engines, agents, and connectors
• Engines — the engine contract, and Hermes
• Architecture — how provisioning and the control plane work
• Self-hosting — run your own cloudrun