Update documentation

Use doxloop update after your product's source has changed, so the agent only inspects what's different instead of re-reading everything.

Prerequisites

• A Doxloop project that has already completed one successful doxloop create

run, so a synchronization baseline exists in .doxloop/sync-state.json.

Run update

Tell Doxloop what changed or what the update should prioritize:

doxloop update "Update affected pages for the latest product changes."

Use a more specific instruction when the update needs special handling:

doxloop update "Document webhook retries and remove the legacy import workflow."

update accepts the same --agent, --model, --reasoning, --reference, --screenshots, --no-screenshots, and --print options as create. See Create documentation.

For screenshot refreshes, use --screenshots --print and submit the prepared prompt in the Codex or Claude desktop app. Capture is not supported through their CLI workflows. See Add guide screenshots.

What happens

Doxloop compares each configured source's current state against the recorded baseline — committed changes since that commit, plus any uncommitted working-tree changes — and includes that comparison in the agent's prompt. The agent:

1. inspects only the reported changed files (not the whole source, when a

baseline exists);

1. finds documentation pages describing affected reader-visible behavior;
2. updates those pages, preserving your confirmed brief, terminology, and

structure unless the change contradicts them;

1. checks adjacent pages for contradictions or newly exposed gaps; and
2. leaves documentation unchanged when a source change has no reader-visible

effect, explaining why.

No baseline yet

If no source has a recorded baseline — a fresh project, or a source that isn't a Git repository — update still runs, but the agent inspects the source directly instead of using a targeted diff. Run doxloop create at least once to establish a baseline for faster future updates.

Next steps