Use this page after the first-run tutorial when creation needs more control. The first-run create --source --output command creates the project and starts authoring in one operation. Inside an existing Doxloop project, doxloop create can extend or improve the agreed documentation set.
Add a first-run request
Pass the audience, outcomes, priorities, terminology, or exclusions with the first-run command:
doxloop create \
--source . \
--output ../your-product-docs \
"Write for platform engineers. Include installation, authentication, and troubleshooting."Extend an existing project
From an existing documentation project, describe the content you want to add or improve:
doxloop create "Write for platform engineers. Include installation, Kubernetes deployment, authentication, and troubleshooting."doxloop create "Prioritize self-hosted deployment and administration. Do not document internal APIs."These instructions work alongside evidence found in your source. Durable decisions get saved to the documentation brief and reused by later create, update, and review runs.
Choose an agent, model, and reasoning effort
Doxloop selects the first supported agent found on PATH. Choose one explicitly, and pass its model:
doxloop create --agent codex "Create a quickstart for developers adopting this product."
doxloop create --agent claude --model <model-name> "Create API documentation for developers integrating this product."
doxloop create --agent gemini "Create an operations guide for administrators running this product."For Codex, set the reasoning effort for the run:
doxloop create --agent codex --reasoning high "Create a production-readiness guide for platform engineers."Supported reasoning values are minimal, low, medium, high, and xhigh. See Agent compatibility for how each agent CLI is invoked.
What the agent does
- Researches your configured source.
- Identifies readers, public capabilities, and important workflows.
- Proposes a prioritized documentation plan.
- Asks once, if a material decision can't be resolved from your request or
the source, combining up to three questions in one message.
- Saves the confirmed documentation brief to
.doxloop/project.json. - Creates or improves the agreed documentation pages and navigation.
- Verifies the result against Doxloop's quality requirements.
docs/ directory (or your generator's content directory) populated with real pages, docs.json navigation updated, and .doxloop/sync-state.json recording a new baseline — but only once validation passes and, for create, a primary audience and priority outcomes were saved.Use a design reference
Add a public documentation site as a presentation and information-architecture reference for this run:
doxloop create --reference https://docs.example.com/ "Use this site as an information-architecture and presentation reference."The reference informs layout, typography, and component patterns only — never your product's claims, examples, or navigation labels. See Use a design reference for capture details.
Work without an installed agent CLI
If no supported agent is on PATH — for example, you prefer an IDE or desktop agent session — print the prepared prompt instead of launching a CLI:
doxloop create --print "Create a quickstart for developers adopting this product."Paste the printed prompt into an agent session opened at the project root. For a new project, include both --source and --output with --print.
Use this path in the Codex or Claude desktop app when an application guide needs screenshots. Screenshot capture is not supported through Codex CLI, Claude Code CLI, or Gemini CLI workflows. See Add guide screenshots.
--print prepares the prompt but can't observe when the external agent finishes, so it does not record a new synchronization baseline. Use the Doxloop-launched workflow when you need automatic source-change tracking for later update runs.