Skip to main content
Capsule uses the same Python entry point for local development and hosted deploys. capsule serve app.py:app starts the local dev loop. capsule deploy app.py:app ships the app. In hosted mode, Capsule runs the app in isolated per-user runtimes so each app user has their own sandboxed execution context.

​
1. Confirm the entry point

capsule serve app.py:app
app.py:app means:
  • module: app.py
  • symbol: app
Serve uploads the project, boots the app runtime, and hot-reloads as you edit watched files.

​
2. Create required resources

Create any resources your app declares before deploy: 
capsule secret create OPENAI_API_KEY=sk-...
capsule fs create reports
capsule channel create support-telegram --type telegram -c bot_token=123:ABC
Then reference them in code: 
app = cpsl.App(
    name="support-ops",
    image=cpsl.Image(python_packages=["openai"]),
    channels=[cpsl.Channel("support-telegram")],
    secrets=["OPENAI_API_KEY"],
    filesystems={"/data": cpsl.FileSystem("reports")},
)

​
3. Deploy

capsule deploy app.py:app
Deploy sends the app definition, runtime image, source archive, pages, integrations, schedules, secrets, filesystems, channels, and product settings to Capsule. The output includes the hosted URL.

​
4. Know what changes in hosted mode

The code stays the same, but requests now come from the hosted app:
  • users sign into this app’s own user pool
  • session.user is the current app user
  • access="authenticated" pages are gated
  • workspace secrets are resolved in the managed runtime
  • each app user gets an isolated runtime for agent work
  • filesystems and container-local files are available to code running in that runtime
  • declared channels can deliver messages to the app

​
5. Iterate

Make changes locally, serve again, then deploy the same entry point when ready. 
capsule serve app.py:app
capsule deploy app.py:app

​
Next steps