CLI

@basemark/cli resolves a markdown(+directives) file into one self-contained .html file — open it in any browser, no server, no build step, nothing else the reader needs to install. This is the third consumption path from the introduction: for when there's no app to embed Basemark into at all, just a document to hand someone.

Install

npx @basemark/cli render doc.md -o doc.html

Runs on plain Node. dist/index.js bundles every dependency in, including @basemark/bio/common/charts.

Usage

basemark render doc.md -o doc.html
basemark skill install [<dir>] [--global]
basemark help

render is the one most people want — one input file, one output file. skill install generates and installs a Claude Skill for authoring Basemark documents, derived from the live component registry: an index of every registered component plus one reference file per component with its full prop details. Installs to .claude/skills/basemark/ by default, ~/.claude/skills/basemark/ with --global, or wherever <dir> points.

What's actually inside the output file

The component runtime and theme are inlined, but only for the domains the document actually resolves a tag from — a common-only document doesn't ship bio's vendor libraries (3Dmol.js, protvista-uniprot, LocusZoom.js), which run into the tens of megabytes uncompressed. The CLI walks the resolved hast tree, works out which domains (common/bio/charts) are actually in use, and inlines only those runtime bundles plus a shared base bundle every document needs.

A bad directive still fails visibly in the output file, exactly like every other render path — a basemark-error element, not a blank space.