Keep your voice. Lose the CMS busywork.

If you write for a living, you know the feeling: the writing goes well right up until it’s time to publish, and then you get pulled into the CMS editor to set the title properly, fix the formatting, add the tags, and wrestle a block or two into place. The writing happened somewhere you love. The finishing happened somewhere you tolerate. And every time you want to fix one clumsy sentence in a published post, you’re back in that web editor again, away from the way you actually like to work.

Specter closes that gap. Connect your Ghost, WordPress, Shopify, or Webflow site in the browser and your published posts open as clean, editable content — title, tags, status, excerpt, and feature image all there, the body as plain prose you can actually read. Edit a post by hand, or run an AI recipe across the whole site and review what it changed. There’s nothing to install, no plugin to configure, and no card-wrangling tax on a two-word fix. The webapp is the fastest way in — subscribe now and you start with 500 free credits when your workspace opens.

AI across your whole archive, not one post at a time

The reason a CMS editor feels small is that it only ever shows you one post. Specter runs AI as recipes inside your workspace — “match the intro style of my last twenty essays,” “fix the internal links so nothing’s orphaned,” “refresh the dates and stale claims” — across everything at once, the way a good editor would. You’re not pasting posts into a chat window one at a time; the whole site is the context.

Browsing and reviewing diffs are free; only AI runs spend credits, and every workspace opens with 500 free — enough to see exactly what it does across your archive. Specter isn’t an AI itself; it’s the workflow and control layer that runs the model over your whole site and shows you every change before it ships.

A faithful projection, not a lossy export

Here’s the part that matters if you care about your posts. Ghost, for instance, doesn’t store your writing as Markdown — under the hood it uses Lexical, the format the Koenig editor draws its cards from. So Specter makes a faithful projection of each post into clean, editable content and carefully translates your edits back on publish. For ordinary writing — paragraphs, headings, lists, links, bold and italic — that round-trips cleanly and you’ll never think about it. Card-heavy posts (galleries, embeds, complex callouts) are where two formats that don’t perfectly overlap can’t always come back byte-for-byte. That’s a property of the conversion, not a bug we’re hiding — and Specter flags it rather than letting you discover it later. How Specter handles Ghost cards walks through exactly what survives.

Review every change before it goes live

Because that translation isn’t always perfectly symmetrical, Specter never just shoves changes live and hopes. Every edit — yours or the AI’s — shows up as a diff first: which posts would change, exactly what would change in each, and anything flagged as a conflict. You approve what ships. If a post changed on both sides since the last sync, Specter pauses and asks rather than silently overwriting one with the other. Every publish keeps a snapshot behind it, so reverting is a decision, not a recovery project.

That combination — a clean copy you can write in, plus a diff that tells you the truth before you commit — is what makes editing a live blog feel safe instead of risky. You’re never guessing what’s about to happen to a published post.

What this looks like day to day

You rewrite a clumsy paragraph in an old post, see the one-paragraph diff, approve it, and it’s live — without ever opening the CMS editor. You ask an AI to bring twelve outdated posts up to date, skim the changes, ship the ten that are right, and send two back. Your archive stays consistent because every change ran through the same review, not through twelve separate trips into a browser admin.

There’s a fair question lurking here: if your CMS has a perfectly good web editor, why route around it? Markdown vs. the web editor makes the honest case, including where the built-in editor is the better choice. And editing with Claude shows what becomes possible once your whole site is something an AI can actually read and revise.

Prefer your content as files on your own disk?

Some writers want the archive as plain .md files on their own machine — in an Obsidian vault, a git repo, wherever — with the engine running locally. That’s exactly what the desktop and open-source edition is for: the same two-way sync, dry-run preview, and conflict handling, running entirely on your Mac with nothing routed through our servers.

The promise is narrow on purpose: keep your voice, edit your published posts without the busywork, let AI help across the whole site, and never publish a change you haven’t seen. Subscribe now →