What you actually own

The claim is that if Reed disappears, your site keeps working. Here is what that means concretely, and how to verify it rather than trust it.

What the build produces

HTML, CSS, JavaScript, images, a search index, and plain-text derivatives. No runtime, no backend, no configuration read at request time.

There is no call to a Reed service anywhere in the output. Search runs in the browser against the pagefind/ index shipped with the pages — not against a hosted search API.

Check it yourself

Do not take the paragraph above on trust. Grep the output:

sh

grep -rl "reedbuild.dev" dist/ | grep -v llms.txt

On a site that does not set site to a Reed domain, the only matches should be links you wrote. There is no telemetry endpoint to find, so if you find a network call you did not author, that is a bug worth reporting.

Build offline, verifiably

The build itself does not need Reed’s infrastructure either. After one reed prefetch, run the build inside a network namespace with no outbound access:

sh

unshare --user --map-root-user --net -- \
  bash -c 'ip link set lo up 2>/dev/null || true; exec reed build --out dist'

A passing build is a measurement, not a claim. This site’s own CI runs exactly this.

What you own, and what you do not

You own: the built site, the generated MCP server source, the generated Skill pack, your docs.json, your content. All of it lives in your repository or your object storage, under your license.

Reed owns: the compiler. If it stops being maintained, you keep every artifact it ever produced and lose the ability to produce new ones from the same source.

That is a real dependency, and stating it plainly is more useful than implying there is none. The mitigation is not a promise — it is that your content is MDX and your config is one documented JSON file, so what you would have to replace is a build step, not a platform.

Hosting it anywhere

A static site with no runtime has no hosting requirements beyond serving files. Cloudflare Pages, Netlify, Vercel, S3 behind a CDN, GitHub Pages, an nginx root, or a directory on a machine you already run.

The two things a host has to get right are the mount path — see Deploy — and the response headers for .md and llms.txt, see Response headers. Everything else is file serving.