We built the pre-publish gate our own pipeline needed. Beta access, 5 slots.

If you're running a pipeline that generates content with an LLM and publishes it without a human approving each post, you already know where it breaks — sensitive data slips in, or the draft violates whatever format the publishing platform expects (frontmatter, tag limits, title characters). Usually both. Someone has to eyeball every post before it goes out.

This API replaces that checkpoint with a deterministic rule set — no LLM judgment involved. Send it one piece of text, get PASS/FAIL and the exact reason back immediately. It's built to sit inside a pipeline with no human approval step — it's the same gate logic our own publishing pipeline runs on every post, every day.

It handles drafts with code examples correctly. Comments and dividers inside code fences are excluded from the checks — developer content is full of code samples, and a naive gate rejects them as format violations.

What it checks

• Sensitive-data keyword/marker scanning (financial-authority limits, approval chains, equity terms, and similar organization-specific categories, plus legal-entity markers). You can pass in your own blocklist at request time for names specific to you.
• Six publish-format axes — no H1 in body, no horizontal rule in body, max 4 tags, no special characters in title, restricted frontmatter keys, and unclosed-code-fence detection.

Beta access — what we're asking

$20/month for 1,000 calls. 5 early-access slots. Not a subscription — monthly prepay, cancel anytime. First 5 people to comment get in. No public price list yet — this is a beta-only rate; the eventual price will be set separately.

The price isn't based on any market benchmark — we don't have one (see below) — it's just what we estimate our own infrastructure margin can absorb. Unconfirmed.

Where this stands — prototype, no billing yet

Right now, access means we hand-issue an API key and track usage by hand. Automated billing (Stripe metered billing or similar) is out of scope for this round — that's a solved problem elsewhere, so we're not rebuilding it until there's demand to justify it.