Operations
Day-two operation of the Compozy daemon — process supervision, SQLite backup and inspection, troubleshooting, and production readiness.
- Audience
- Operators running durable agent work
- Focus
- Operations guidance shaped for scanability, day-two clarity, and operator context.
This section is for keeping Compozy running. It covers daemon supervision, database state, the diagnostic flow when something misbehaves, and the production readiness checklist before Compozy runs unattended in a real environment.
Use it after the happy path works. Getting started pages show how to launch the daemon; operations pages show how to prove it is healthy, where state lives, what to back up, and which checks matter before a long-running agent run depends on Compozy. Most operators should read the daemon page once, then keep troubleshooting and production readiness nearby for incidents and release preparation.
The database page is intentionally separate from the daemon page because Compozy has two persistence layers: the global catalog and per-session event stores. When a session looks wrong in the UI or CLI, check daemon health first, then inspect the relevant database path with the backup and recovery rules from this section.
In this section
Daemon operations
Use this page when starting, stopping, restarting, or supervising Compozy as a long-lived process.
Expose Compozy over MCP
Connect a trusted MCP client to one workspace over stdio or authenticated loopback HTTP.
Database operations
Use this page when backing up SQLite, inspecting stored sessions, or retiring session directories.
Support bundles
Use this page when creating a redacted daemon-owned bundle for maintainers or incident triage.
Troubleshooting
Use this page when an operational command fails or the daemon misbehaves.
Production checklist
Use this page before running Compozy unattended in a real environment.