Deploy contract
Review the manifest like production code.
Application code and deploy configuration change together. A small manifest keeps the operational parts visible in source review instead of scattering them across a chat transcript, dashboard, and shell history.
| Field | What to check |
|---|---|
| name | Stable app name used by Adios for the workload. |
| build_cmd | Command that prepares production assets from a clean checkout. |
| start_cmd | Command that starts the production process. |
| runtime.port | Port the app listens on inside the runtime. |
| runtime.health_path | HTTP path Adios can use to test readiness. |
| secrets | Secret references such as secret://DATABASE_URL. |
| requires | Managed services the app depends on, such as db or cache. |
Workflow
Use the manifest to keep AI output deployable.
When an AI agent adds an API route, background worker, database call, or external provider, ask it to update the manifest only when the runtime contract changes. Then verify the command and config instead of trusting the generated file blindly.
Build from a clean tree
The build command should work without editor state, local caches, or undeclared tools.
Start the production server
The start command should run the server process that will receive real traffic.
Probe readiness
The health path should fail when required request-path dependencies are unavailable.
Choose a deployment path
Review framework, runtime, service, and workflow requirements before writing the manifest.
Manifest reference
Read the complete field reference for app and workflow manifests.
Quickstart
Deploy an existing app from your terminal with the user CLI.
Deploy AI-generated code
Connect generated source changes to previews, logs, and production.