How it fits together
A one-line installer adds the Monk plugin, or amonk MCP server entry, to the coding agents on your computer: Claude Code, Codex, Cursor and others. Your agent talks to Monk’s local companion, monk-agent, which talks to Monk’s orchestrator. Your app’s resources run in your cloud account on AWS, Google Cloud, Azure, DigitalOcean or Hetzner, or on your own machine.
Quickstart
Install, sign in, ask for a plan and approve it
Two ways to start
I’m starting a new app. Build it with your coding agent, then ask Monk to plan a deployment. You review the plan and its estimated cost before anything is created. My app is already running. Monk builds a new environment from your source code alongside the current one and leaves the original alone. It doesn’t copy your data, undocumented settings or external account state. Moving data and switching traffic are separate steps you review. Both environments cost money while they run side by side. Either way, the first prompt is the same:What you do and what Monk does
Monk doesn’t edit your application code. Plans, credential forms and other approvals open in the local dashboard on your computer, so credentials never go into the chat.
After the first deployment
- The orchestrator restarts and reschedules managed workloads within its policies.
- Watcher, if you set it up, sends crash and resource alerts with an AI-written diagnosis to Slack. Any fix it proposes waits for your approval.
- CI/CD with GitHub Actions deploys when you push.
- You can ask for status, logs and cost at any time.
Where to go next
Quickstart
From install to an approved plan
First deployment
What each approval shows
How Monk compares
Monk next to Terraform, Kubernetes and hosting platforms
Integrations
Services and providers Monk can manage

