Services and databases
Ask Claude to keep an app, worker, database, or other supporting process available.
Most useful projects need more than source files. They may need a web application, Postgres, a queue worker, a local API, or a process that turns incoming data into a report.
Claude can set up those pieces inside the pod, connect them, verify them, and mark the important ones to return after a crash or restart.
Ask for the working system
Describe what should be true when the work is finished:
Run the application with Postgres and the queue worker. Apply the migrations, verify one job from end to end, and keep all three available after the pod restarts.
Turn this script into a small private dashboard. Refresh its data each morning, keep the history in Postgres, and send me the link.
Keep the development server running while we work. If it stops, recover it and preserve enough logs to explain why.
This gives Claude room to choose the right project commands while making the durability and verification requirements explicit.
What Podbay can keep available
- the project's main development server;
- workers, queues, watchers, and local APIs;
- Postgres or another project-local data service;
- a bridge or helper required by an internal tool;
- a private dashboard that exposes the result.
Prepared environments may already define the main application service. Claude can add other long-running project services when needed. Podbay relaunches declared services after a pod restart and watches for common failures while the pod is running.
Services are different from scheduled work
A service stays available and waits for traffic or work. A schedule wakes Claude at a particular time to complete a task. Many projects use both: a dashboard stays running while Claude refreshes its data each morning.
Decide where important data belongs
Project-local Postgres and persistent files work well for development, prototypes, internal tools, and small personal workflows. Ask Claude to keep their data in the persistent workspace and to document how it is backed up or exported.
Move to an independently managed database or service when you need production availability, independent backups, compliance controls, or scaling beyond one pod.
Verification matters
“Keep the worker running” is incomplete without a way to know it works. Ask Claude to exercise a real job or health check and to explain where failures will be visible.