QUILLRUN
New · Connect Claude, Cursor or ChatGPT to your own workspace

The autonomous SEO content agent that publishes straight to your CMS.

Quillrun researches from real sources, drafts against the facts it found, runs every piece through quality gates, holds for your approval, then publishes to WordPress, Shopify or Webflow. One button stops everything.

Free plan · connect a site in about a minute · nothing publishes without your say-so
example run · your-site.com
01Research — 12 sources fetcheddone
02Facts extracted to knowledge basedone
03Draft written against those factsdone
04Quality gates — 5 of 5 passeddone
05Waiting for your approval4h
06Publish to WordPressqueued
◎ Holding at approval gate — 1 draft waiting on you
Publishes intoWordPressShopifyWebflow·Reads fromSearch ConsoleGoogle AdsPlaceholder — send customer logos

Six steps, and you can stand on the brakes at any one of them.

The whole loop is visible while it runs. Two of these steps are new this quarter — research is now grounded in fetched sources, and the agent tells you what to fix after publishing.

01RESEARCHNEW

Pulls your Search Console queries and Google Ads volumes, fetches real sources, and extracts the facts into a knowledge base for your site.

02DRAFT

Writes against those extracted facts rather than from model memory, so claims trace back to a page it actually read.

03QUALITY GATES

Length, keyword coverage, internal links, originality, citability. Miss one and the run stops — it fails closed, never open.

04APPROVAL

Optional human gate. Leave it on and nothing touches your site until you press approve; turn it off per site once you trust it.

05PUBLISH

Into WordPress, Shopify or Webflow as a real post or product page. Three failed publishes in a row and the site auto-pauses.

06RECOMMENDNEW

Watches indexing and traction afterwards, then ranks what to fix or refresh — closing the loop instead of just adding pages.

NEW

Research that cites its sources

The agent fetches real pages, extracts facts into a per-site knowledge base, and drafts against those facts. Every claim in a draft traces back to a URL it actually read — which is also what makes the writing citable by AI answers.

Quality gates that fail closed

Length, keyword coverage, internal links, originality, citability. A draft that misses a gate never reaches your site — it stops and tells you which gate it failed and why.

✓ LENGTH✓ COVERAGE✕ THIN SECTION
NEW

It tells you what to fix next

The recommendation engine watches Search Console — title and meta problems, keyword gaps, indexing errors, pages with zero traction — and ranks what to fix on your existing pages, not just what to write next.

New · MCP server

Bring your own AI client.

Generate an API key and point Claude, Cursor, ChatGPT or Codex at your workspace over MCP. Your agent reads your sites, drafts, runs the same gates and publishes — scoped to your org only, with per-key monthly limits and one-click revoke.

Read the MCP docs
mcp.json
"quillrun": {
  "url": "https://quillrun-api.vercel.app/mcp",
  "headers": {
    "Authorization": "Bearer qr_live_ab12…"
  }
}

You are still in charge of the thing.

Every objection to letting an agent write on your domain has a control here.

One kill switch

Stop every run across every site from the bar that sits above every screen.

Approval gates

Per site. Nothing publishes without a human until you decide otherwise.

Auto-pause

Three consecutive publish failures pauses the site rather than hammering your CMS.

Full audit log

Who changed what, which key did it, when. Exportable for a client review.

Pick the version of this that's yours.

Team of oneNo SEO hire, still needs to publish weekly

Set a schedule, approve on your phone, let the gates do the reviewing you don't have time for.

Talk to us →
AgencyTwelve clients, twelve CMSes, one dashboard

One org per client, separate keys and limits, an audit log you can hand over in a review.

Talk to us →
Ecommerce900 product pages, none of them written well

Optimize product copy in batches, ranked by which pages actually get impressions today.

Talk to us →
2posts published
29%runs that succeed
3CMSes, one API

Questions people actually ask

Will this publish rubbish on my site?It can't reach your site unless it passes every quality gate, and by default a human approves each piece. Both are visible while a run is in flight.
Does Google penalise AI content?Google ranks helpful content regardless of how it was produced. The gates exist to enforce the "helpful" part — grounding, originality, depth — not to hide the origin.
Which CMSes do you publish into?WordPress, Shopify and Webflow today, over the CMS's own API. Search Console and Google Ads are read-only inputs.
Can I use my own AI client instead of yours?Yes — generate an API key and connect Claude, Cursor, ChatGPT or Codex to your workspace over MCP. The same gates and limits apply.
What happens if I want to stop everything right now?One button in the status bar halts every run and blocks publishing org-wide until you resume.
How do agencies keep clients separate?One org per client, with its own sites, keys, limits and audit log. Keys are scoped to a single org and revocable in one click.

Connect one site. Watch one run. Then decide.

Free plan, no card, and nothing reaches your site until you approve it.

Start freeTalk to us