The brand book your agents can actually use.

An open standard that packages brand knowledge as plain Markdown. Easy to maintain, written for agents and efficient context usage.

BRANDBOOK.md
---
standard: brandbook.md
spec_version: "0.1"
profile: evidence-based # optional
brand:
name: Acme Hot Sauce
files:
- path: assets/dna-inventory.md
summary: "Assets and usage rules"
- path: governance/anchor-rules.md
summary: "Non-negotiables"
---

Why now

The next generation of the brand book

Good brand books left the PDF behind years ago. From the printed manual to the PDF to today's brand portals, every generation made the rules easier for people to reach.

AI agents write copy, generate imagery, and answer customer questions as your brand, and a portal can't tell them the rules. The next step is a brand book that is data: versioned like software, open to review, and readable by any agent your team already works with.

Printed manual

Bound rules on a shelf. Authoritative, unreachable.

PDF brand book

Searchable, emailable — a picture of the rules, not the rules.

Brand portal

Live and linkable, locked behind a human-first interface.

brandbook.md

Versioned like software, open to review, readable by any agent.

The shape of it

A root file, and the whole structure underneath.

The brand book becomes a folder of structured and versioned files. Using YAML and Markdown standards, all content is reachable and processable, without overloading the models' context. The structure remains flexible, so your brand, as well as your brand definition framework, can be fully represented.

EXPLORER · brand/
assets/
voice/
audience/
context/
governance/
brand/BRANDBOOK.md ROOT — ALWAYS LOADED
---
standard: brandbook.md
spec_version: "0.1"
profile: evidence-based # optional
brand:
name: Acme Hot Sauce
category: "Hot sauce / condiments"
files:
- path: assets/dna-inventory.md
summary: "Assets and usage rules"
- path: voice/voice.md
summary: "Tone, vocabulary, examples"
- path: audience/buyers.md
summary: "Buyer personas"
- path: context/history.md
summary: "Origin and milestones"
- path: governance/anchor-rules.md
summary: "Non-negotiables"
---
# Acme Hot Sauce
The root loads first. Everything below it
is fetched on demand via the files index.

Why files, not one file

Load less, know more.

Because the book is split into files behind one index, an agent's context carries only the rules the task needs. Files and sections can reference each other.

The usual way
One megafile

The entire book travels in every request — relevant or not.

strategy in context assets in context voice in context audience in context governance in context
CONTEXT USED: ALL OF IT, EVERY TIME
BRANDBOOK.MD
A package behind an index

The compact root, plus only the file this task needs.

root always voice this task assets skipped audience skipped governance skipped
CONTEXT USED: A FRACTION, ALWAYS FRESH

Smaller prompts

Only the relevant rules travel, which means less context & lower cost on every single call.

More room to think

Context spent on the task instead of boilerplate means sharper, more on-brand output.

Fresh rules

Placed centrally with MCPs or direct access, to make every agent pick it up on the next load. So no re-export, no stale PDF.

The principle

The most open standard for brand knowledge.

Your brand book is as unique as your brand. That's why we define as much as needed and as little as possible, so every brand fits: CEP-driven, persona-driven, or technical down to the design tokens.

01

Works with every brand.

The system is given; the methodology is yours. Personas or category buyers, archetypes or distinctive assets: file types are open, agents work with whatever your book carries.

02

Profiles for a quick start.

We work on providing methodology profiles that give you a head start. Think of it as section templates based on existing brand definition frameworks, that are already known by LLMs.

03

AI-native data structure

Versioning, review cycles, and numbered anchor rules keep the book current and citable. Agents can process your contents directly, without needing a handbook about the brandbook standard itself.

Reference packages

Examples: Two approaches on the same standard

Both conform to the same standard, because every brand book is allowed to be different.

Acme Hot Sauce
PROFILE: EVIDENCE-BASED

Multi-market brand with a sub-brand hierarchy, built on the evidence-based profile.

Read →
Lumen Coffee
PROFILE: NONE — BASE STANDARD

A personas-and-archetype brand that declares no profile — the base standard alone.

Read →

The landscape

There are multiple other approaches. Find yours.

The current pace in model development, new tools & new approaches is insane, and AI FOMO is real. Some smart ideas might be outdated before you even get started.

We believe that the only solution is openness and collaboration. We build brandbook.md with that in mind. We want it to work with every AI tool & every model. We want it to work today and tomorrow. And if not, we want it to be open and easily transformable into whatever the future of AI, marketing & branding brings.

Read the full comparison — including when it's not us →
Built in the open
Text CC-BY-4.0 · Code MIT · Draft 0.1 — feedback shapes it right now.