> ## Documentation Index
> Fetch the complete documentation index at: https://docs.crewship.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Find your way around Crewship

> Choose a starting page by the job you want to accomplish instead of by the command or API resource name.

# I want to…

| I want to…                                           | Start here                                                                |
| ---------------------------------------------------- | ------------------------------------------------------------------------- |
| Run my first agent task                              | [First projects](/guides/first-projects)                                  |
| Understand crews, agents, and Keeper                 | [Concepts and glossary](/concepts)                                        |
| Give an agent a task that produces a reliable result | [Writing a good task](/guides/writing-good-tasks)                         |
| Automate recurring work                              | [Routines](/guides/routines) and [scheduling](/guides/scheduling)         |
| Give an agent access to an API key safely            | [Credentials](/guides/credentials)                                        |
| See what a run changed                               | [Crew Journal](/guides/crew-journal)                                      |
| Describe a workspace declaratively                   | [Manifest guide](/guides/manifests)                                       |
| Decide whether Crewship fits the job                 | [When not to use Crewship](/guides/when-not-to-use)                       |
| Find an exact HTTP route or CLI flag                 | [API reference](/api-reference/overview) or [CLI overview](/cli/overview) |

## Pick the smallest useful path

Start with a direct agent task when you are still learning the workspace and want to inspect one result. Move to a routine when the task has a repeatable definition and a trigger. Add an issue when work needs ownership, status, or a reviewable queue. Add a waitpoint when an external decision must pause execution without losing the run context.

For access questions, read [Concepts](/concepts) first: a workspace owns the tenant-wide boundary, a crew owns the shared runtime, and an agent is the worker. For safety questions, read [Credentials](/guides/credentials), [Keeper](/guides/keeper), and [Harbormaster](/guides/harbormaster) together; no single page describes the entire access decision.