BRAND.md and DESIGN.md: your brand in two files every AI tool reads
Every AI tool starts from zero about your brand: who you are and how you look. brand.md is the open standard that puts an end to that, with two files from one scan. The generator for it is free.
You draft a newsletter in ChatGPT, have Claude sketch a campaign, ask an image model for a visual, and a colleague answers email with an AI assistant. Four tools, four times explaining who you are and how you look. And because no one gives that explanation the same way twice, your brand sounds and looks slightly different in every channel.
That’s not a time problem, it’s a consistency problem. The explanation lives in one person’s chat history, not in the brand. The moment that person is on holiday, the next one starts from zero again.
What brand.md is
brand.md is an open standard: plain text, readable by a human and by a language model. You put the files in the root of your project, next to README.md, or you paste the content into a chat. No integration is needed: any model that reads text reads them.
The format isn’t ours. brand.md is an open standard under the MIT licence, created by Caio Pizzol; the spec is at version 0.3.0 and public on GitHub. We built the tooling around it. GitHub.
The file belongs in a lineup you might already know:
llms.txt
Tells a language model which pages on your website matter and where they live.
AGENTS.md
Tells a coding agent how your project is put together: commands, conventions, boundaries.
BRAND.md + DESIGN.md
Tells every AI tool who you are and how you look: what you stand for, how you sound, and which colours and typography go with it.
One scan, two files
BRAND.md says who your brand is: voice, audience, guardrails. DESIGN.md says how it looks: colors, typography, sizing. The open standard draws that line itself; we deliver both sides from the same source. An agent that only writes text has enough with BRAND.md; an agent that also builds a UI, slide or image needs both.
What’s in BRAND.md
The standard prescribes Strategy and Voice. Branddock adds Audience, Products and Guardrails, which are needed the moment an agent does real work for you. Those extensions are additive: a tool that doesn't know them skips them and is left with a valid file.
Strategy
standardWhat the brand stands for: positioning, personality, promise, audience and guardrails.
Voice
standardHow the brand sounds: tone, words you do and don’t use, example sentences from your own copy.
Audience
BranddockYour personas as separate blocks, so an agent can write for the right reader.
Products & Services
BranddockWhat you offer, with the benefits and use cases per product.
Guardrails
BranddockDo’s and don’ts as machine-readable lists, so a tool can check them instead of interpreting them.
What’s in DESIGN.md
The applied visual system: not how the brand sounds, but how it looks. The same additive rule applies: a tool that does not know a section skips it.
Colors
Brand colors with their hex values and semantic role (primary, surface, on-surface), so no one has to guess which color is for what anymore.
Typography
Typefaces per role (headings, body text) with size, weight and line height.
Layout & Components
Sizing, rounding and the built-up style of reusable parts like buttons, so a generated UI isn't styled off the cuff.
Curious what will show up for your brand?
Paste your URL and get your own BRAND.md and DESIGN.md, with each section honestly marked as confirmed or not by the scan.
Generate your brand.mdFree · no account, no email needed for the files
Why we embrace an existing standard
We could have launched our own format. We deliberately didn’t, and the reason is in two recent examples.
llms.txt showed how it goes wrong. Many sites published one, and that counted as adoption. Until someone looked at the reading side: the major crawlers barely do anything with it. A format with only writers and no readers is a dead letter.
MCP showed how it goes right. That protocol didn’t win by recruiting writers, but by shipping the reading side on day one: Claude read it immediately. The rest followed.
Two competing formats under the same name would both lose. So we adopt the existing spec as the core, publish our extensions as an addition, and submit the generally useful pieces upstream. What we add is the tooling: a generator, a validator and a living version, for both files.
What a scan can’t know
A scan of your website can't verify anything. It reads what's there and infers the rest. That's exactly where a generator tends to start lying convincingly, so these are the three rules we stick to, for BRAND.md and DESIGN.md:
The file says whether a section is filled, not whether it’s correct
Every section gets filled or empty in the frontmatter: purely whether there's content. Only a human can confirm that content is correct, and we don't claim that anywhere in the file.
Missing stays missing
A required section with no data says “Not yet defined” instead of a convincing-sounding paragraph. In our own file that’s currently the case for Positioning. We leave it as is.
The Brand Score is explainable
You get a score with three explainable checks: completeness, consistency and machine readability. No AI handing out a number nobody can account for.
Free files, paid upkeep
The generator is free and stays free. You paste your URL, you get your BRAND.md and DESIGN.md, right away and without an account or email address. You see your Brand Score immediately; for the dimension breakdown and the findings per section (the full report) we ask for your email address. You also get that report by email, with your download link in it, so you can pick it up later. We'd rather say that here than have you discover it along the way.
What we sell is that the files stay current. A downloaded file goes stale silently: you reposition, your look changes, you launch a product, and the files your agents read know nothing about it. In Branddock, your BRAND.md and DESIGN.md are a living version with a fixed URL, that moves with your brand DNA and that your agents always fetch fresh via the MCP connector. The `validation:` and `provenance:` fields in the spec can only take on real meaning there: they say who confirmed what, and when, and where the current version lives. MCP connector.
That’s the whole model. The format is open, the files are yours, and you don’t need us to use them.
What brand.md doesn’t solve
BRAND.md fixes how your brand sounds, DESIGN.md how it looks. Neither decides whether a campaign is a good idea, whether a message is right, or whether the moment is right. That stays human work, and doesn’t change as the files get better.
The files also go stale silently. If your brand changes and the files don’t, every tool keeps writing and styling in the tone and colours you had a year ago. That’s not a flaw in the format; it’s the reason upkeep is part of the deal.
Frequently asked questions
What is brand.md?
A set of markdown files that describe your brand in a fixed structure, so a language model can read them before writing or designing. It’s an open standard under the MIT licence; the spec is at version 0.3.0.
What’s the difference between BRAND.md and DESIGN.md?
BRAND.md says who your brand is: voice, audience, guardrails. DESIGN.md says how it looks: colors, typography, sizing. An agent that only writes text has enough with BRAND.md; if it also builds a UI, slide or visual, it needs DESIGN.md too. Both come from the same scan.
What’s the difference with AGENTS.md and llms.txt?
All three do something different. AGENTS.md tells a coding agent how your project works. llms.txt tells a language model which pages on your site matter. brand.md tells every AI tool how your brand sounds, thinks and looks. They’re not competitors and can coexist.
What belongs in the files?
BRAND.md asks for Strategy and Voice; DESIGN.md the applied visual system (colours, typography, spacing). The most steering part of BRAND.md is concrete example sentences: a model learns faster from a good and a bad sentence side by side than from an abstract instruction like “be authentic”.
Do I need Branddock to use brand.md?
No. The format is open, the files are yours, and you can write them by hand or have another tool generate them. Our generator takes the blank page off your hands; our subscription keeps them current.
What does the generator cost?
No money and no account. Downloading both files is instant, no email needed. We do ask for your email for the full report behind it: that’s before you see the per-dimension breakdown. You also get it by email, with your download link, so you can pick it up later.
How do the files stay current?
That’s the weak spot of any downloaded file: it doesn’t change as your brand does. Put a recurring check on your calendar, or let the living version in Branddock move with your brand DNA.
Free · both files download instantly
28 days free · 300 credits · no credit card