moshcode.sh

The home page for moshcode — a metal wrapper CLI for agentic coding.

Before this repo, moshcode.sh was a parked domain: it rendered as a tenant config inside the moshcoding Next.js app (configs/moshcode.sh.json), the same way moshscript.com and the rest of the parked names do. This is the real site, on its own Railway service, so the product domain stops sharing a codebase with the parking lot.

Run it

npm install
npm start           # http://localhost:3000
npm run dev         # same, with --watch
npm test            # node --test

What it serves

route
/	the landing page
/commands	all 51 commands, grouped
/commands.json	the same table as data
/blog	the post index
/blog/<slug>	one post
/blog/rss.xml	the feed (/blog/feed.xml is the same thing)
/llms.txt	a plain-text brief for agents
/install.sh	302 to https://moshcoding.com/install.sh
/healthz	Railway's health check
/robots.txt, /sitemap.xml	the usual

/install.sh is a redirect, on purpose

There are two installers in this world and they are not the same script: moshcoding.com/install.sh installs into $MOSHCODE_HOME/pkg, while the raw install.sh in the CLI repo replaces ~/.moshcode wholesale. Serving a copy here would make a third one to keep in sync, and the first time it drifted someone would lose a config directory. So the route redirects and curl -fsSL follows it.

The command table is generated

data/commands.json is built from the <!-- COMMANDS --> block in the moshcode README, which the CLI itself emits with moshcode help --markdown and guards with a test. So the site cannot advertise a verb that does not exist, or miss one that does.

npm run sync:commands                               # pull from GitHub
MOSHCODE_README=../moshcode/README.md npm run sync:commands   # or a local checkout
npm run check                                       # exit 1 if it is stale

Re-run it after a moshcode release that adds or renames a command.

Content

All copy lives in src/content.mjs — hero, sections, terminal panes and footer links, as data. src/site.mjs is layout only, so a wording change never means touching markup. Every terminal pane on the page is real output from the moshcode README rather than an invented screenshot.

The blog

Posts live in src/posts.mjs, one object each, with the body as an ordered list of blocks (h2, p, list, terminal). src/blog.mjs renders the index, the post page and the RSS feed from that array using the same chrome as the rest of the site. Adding an entry is the whole publish: the index, the feed, the sitemap and the canonical URL all come off it, and a test asserts every post is reachable and carries its own canonical tag.

Paragraph and list text is authored markup and is passed through unescaped so a sentence can carry a link. Terminal panes are escaped, so pasted output stays verbatim and inert.

Deploy

Railway, from Dockerfile, with railway.json setting the health check. Pushing to main is the whole deploy.

MIT.