Knowledge base
How Work on Repeat thinks about routines, runs, permissions, and failure.
The guides show you what to do. This section explains why the product behaves the way it does, so you can predict it rather than discover it.
Concepts
How Work on Repeat works
The objects, the execution path, and where each guarantee comes from.
Organizations and projects
Who you work with, versus what you are working on.
Routines
Instructions, schedule, limits, lifecycle, and what activation means.
Runs and traces
What is recorded for every execution, and what is redacted.
Run lifecycle
Queued through terminal, and who moves a run between states.
Connections
Managed providers, self-managed MCP servers, and credential handling.
Tool permissions
Disabled by default, enabled, and automatic — three distinct states.
Models
One directory and one inference path, through LLMGateway.
Working well
Writing instructions
A standing responsibility is not a prompt. What changes, and why.
Variables
Substitution rules, naming, and what happens to a missing key.
Schedules and timezones
Five-field cron in an IANA zone, including daylight saving.
Security model
Isolation, secrets, bounded execution, and outbound network rules.