RankMill is an agent that writes your blog. Not a chat window you paste prompts into, and not a template filler. You give it your website, your competitors, and the rules your copy has to follow, and it comes back with topics worth writing, drafts written in your voice, and an editor you can argue with until the post is right.
We built it because the hard part of content was never the typing. It was everything around the typing: deciding what to write, remembering what you already published, keeping eight posts sounding like one company, and doing that again next week.
What it actually does
The loop has three agents and one human, and the human is always the one who says yes.
- Research reads your site and your competitors, builds a written profile of the business, and proposes topics with a reason attached to each one.
- Writer takes a topic you approved and produces a finished markdown post, constrained by the writing rules on that project.
- Editor is a conversation. You tell it what is wrong, it revises the post, and it remembers the whole thread, so the fourth round of notes still knows what the first round asked for.
Nothing moves without you. A suggested topic sits there until you approve it. A draft stays a draft until you say ship it.
The website profile is the thing that makes it sound like you
Most AI writing tools start cold on every request, which is why every draft reads like a slightly different company wrote it. The first research run on a project writes a markdown profile of your business: what you sell, who buys it, how you talk about it. Every later agent reads that profile before it writes a word.
A generic model with a specific brief beats a fine-tuned model with no brief. The brief is the product.
You can edit that profile by hand, or ask an AI to revise it and review the diff before it applies. It is a document, not a hidden embedding, so you can always read exactly what the agents believe about you.
The parts people ask about first
Writing rules. Per project, not per account. "Never say leverage", "second person", "no more than one exclamation mark per post". They are handed to every agent that produces copy.
A schedule. Research can run on its own, weekly by default, at an hour you pick. New topics show up without anyone remembering to ask for them.
Import. You already have posts. Bring one in verbatim, have the importer clean up its structure, or have the rewriter keep every fact and say it again in your voice.
Audit. Point it at a project and it reads everything published, lists what is wrong, and fixes the issues you pick.
A run budget. Every agent execution is one run against a monthly cap, visible on the Activity page with what it did and how long it took. No surprise bills, no black box.
What it is not
It is not a publish button. RankMill does not own your CMS, does not have your credentials, and does not push anything live. It produces markdown, and you take it from there, by download, by copy, or with the CLI.
It is also not a volume play. Ten posts you would sign your name to beat a hundred that fill a sitemap. The cap is a feature.
Start with one project
Point it at your site, let research run, and approve one topic. You will know within about ten minutes whether the drafts are close enough to be worth editing, which is the only question that matters.
That is the whole pitch. Welcome to RankMill.