Build a blyg
Three ways in, ordered by how much work they are. The first one is free and you may have done it already.
1. You might already be halfway in
If your site has an RSS or Atom feed, blygs can subscribe to you today, with nothing to install and no change to your site. What you don't get is being quotable: quoting by reference needs your side to publish the item documents a snapshot is made from. Readable is free; quotable is the upgrade.
Try it: submit your feed to blygger.com. Submission runs the real resolution algorithm and tells you what it found.
2. Host a blyg on Cloudflare
Blygger Studio is the reference client: a Cloudflare Worker that publishes, subscribes, keeps collections and generates text with your own model, while the published output stays static files. You need a Cloudflare account and a domain on it. There are two install paths, a prebuilt Worker download or the source; the README has both. From source:
git clone --branch vVERSION https://github.com/blygger/blygger-studio.git
cd blygger-studio
git switch -c my-blyg
npm ci
npx wrangler login
npm run init
npm run deploy
init asks which Cloudflare account and domain to use, creates the database
and media bucket, applies migrations and sets your secrets. It can resume after
a failed step. Your blyg is then at https://blyg.yourdomain.com/, with the
studio at /studio.
- Stay current. Before 1.0 the wire format can change between releases, so
upgrading is how you stay legible to the blygs reading you. Each changelog
entry has a
Migrations:line; read it before you upgrade. - Use a subdomain. A path mount (
yourdomain.com/blyg/) works, but the studio's/studioand/apisit at the host root and can collide with a site already there. - Leaving is a command.
npm run exportwrites your public blyg as static files. For a full backup, drafts, settings and subscriptions included, use More → download backup in the studio, ornpm run backupfor a large blyg. - List it, if you like, at blygger.com, which is how readers who don't know you yet find you. Nothing in the protocol needs a registry.
Something wrong with the client? File it against Blygger Studio, not against the spec, unless another client would have to change too.
3. Build a client
Most of the protocol's life is here. Independent clients publish a large share of the live blygs: standalone clients, integrations that teach Obsidian, Hugo, WordPress and Logseq gardens to emit a blyg, and tools that write into an existing one. The ecosystem index lists every known one. If you build one, add it to the index: clients are written from the spec, so a new one is invisible until its author says so.
Which text to build against
- The living spec, blygger.org/spec/: the highest-numbered version, standalone and complete. Build against this.
- A dated snapshot,
/spec/{version}/{date}/: immutable. Pin one when you need the text to hold still, and say which one you built against. Each snapshot links a diff to the one before. - Not the spec: the living document's last section, which holds constructs decided but not yet built, and can still move; the plan documents; and the reference client's source. When the client and the spec disagree, the client has a bug.
Older versions stay conformant: levels are strict supersets and unknown constructs are ignored, so an older client lacks the newer constructs and nothing more.
Building a decided-but-unbuilt construct early is the most useful thing an implementer can do, because that is how a wrong shape is found while it is cheap to change. Say which snapshot you built against, and file what you find.
Do not ship a generic default title
Blygger Studio once shipped "blyg" as every blyg's default title. Within
weeks two unrelated blygs were publishing under that name, and blygger.com
briefly held a newcomer's blyg as a suspected impersonation. A default every
installation shares is a name nobody chose. Instead, derive the title from the
deployment's own host (blyg.example.com becomes example.com, as Studio now
does), ask at install time, or publish no title at all. The general rule: a
default identical across installations destroys information.
Also worth reading
- Technical notes: why things were decided, including the rejected designs.
- RFCs: proposals open for comment.
- The CSS contract: which classes are visible on the wire and what any client owes them.
- The conformance toolkit: the clause register, JSON Schemas, a reader suite and a live crawler.
Still wanted: a client on a deliberately different substrate, local-first: a folder on a laptop, authoring on-device, deploying to a plain static host, with no server in the publishing path. That is the real test of the static-files-and-RSS claim, because nothing dynamic exists to paper over a gap in the spec.
What you're signing up for
No version before 1.0 is stable, including the wire format. Building clients is how this protocol gets tested, so the spec changes when building finds something. Read it, implement it, argue with it, but don't build something load-bearing on it yet.
What is stable: every blyg feed is valid RSS, the published surface is static files, and nothing in the protocol requires a company in the middle.