The Brand Context Protocol.

Every brand is quietly becoming a system, not just a look. Most haven't noticed yet.

A brand today has to show up across 50 surfaces, 67 tools, hundreds of little agents, and each one wants structured input before it will produce anything. The PDF brand book can't answer them. The design system answers a few, the website a subset, the AI stack a third, all slightly differently. None of them talk to each other, and what used to look like a minor inconsistency now comes across as a brand contradicting itself in real time. AI didn't create that problem. It just removed the friction that was politely hiding it.

The Brand Context Protocol is our answer: the one source every surface draws from, and the one source every output is checked against. We're releasing it now because the problem it solves stopped being niche.

A brand inside someone else's tool is rented, not owned.

The market is filling up with closed platforms that promise to make your brand "AI-ready." They sign you onto their database, their schema, their roadmap. The pitch usually leads with a number, like cutting agency spend by half. The catch is more subtle: a brand that lives inside a vendor's tool isn't an asset you own, it's one you rent until the tool stops fitting. By then leaving is somewhere between painful and impossible.

There's a second problem underneath. Left to its own devices, AI gives you back the average of everything it has already seen. Competent enough, and instantly forgettable. That's the exact opposite of what a distinctive brand needs. Volume stopped being the hard part. Knowing which output is right, and refusing the ones that are merely close, is the hard part. That judgment has to live somewhere a machine can read, or it never scales past the one person holding it in their head.

So we start from one commitment: the brand stays with you. Always. Whatever we build, you can fork it, take it in-house, or run it somewhere else tomorrow. It's plain markdown sitting in your git, so there's no login to get into and nothing to escape from later.

What it is, in plain terms.

The Brand Context Protocol is a way of writing a brand down as something usable, by people and by machines. Not a SaaS you log into; a small, portable, git-native standard for everything that makes a brand itself, in a form both a person and an AI agent can read and produce from.

Underneath, it's markdown and a little YAML in a repository, with tiny bits of code only where a word can't do the job (a motion curve, a grid, a token). Human-readable, machine-queryable, versioned, yours. Honestly, someone could copy the whole structure in a weekend, and we mean that as a feature. The format is deliberately boring. What's hard to copy is the taste: deciding what goes in, what gets refused, and how lessons from real work flow back to sharpen it.

Here's roughly what one looks like, sitting in your repo:

We're open-sourcing the format.

The structure of the Brand Context Protocol is open, and we're giving it away on purpose. A standard only earns the name if anyone can pick it up and use it, and we think this is how brands should be written for an AI era.

It also reflects where the real work is. Anyone can take the structure. What separates one brand from another is the content poured into it, and the taste behind every call about what belongs and what does not. Brad Frost's Atomic Design works the same way: the method is there for everyone, and what you make of it is entirely yours. We give the structure away and bring the judgment.

Brand Truth is the single source: everything we know about the brand, with a source and a reason behind each claim. Voice, design, the rules, the beliefs, and the refusals (the things the brand will never do, no matter who asks). It's the place we build from. Truths get added and dated, and the old ones are kept rather than quietly overwritten.

Skills are small reusable tools that capture a process: a landing page, a deck, an ad, an email. Each one is a contract. It names what it needs, pulls the right parts of Brand Truth, and produces the work. One skill might scale a campaign across every audience and format, another might check any asset against your standards. Each new skill is a permanent capability, so the system gets more capable the longer you run it.

Output is what comes out: pages, decks, ads, email, motion, generated from Brand Truth rather than from a blank prompt at 11pm. Because it can only draw from the brand system, it is on-brand the moment it is made, and every asset arrives with a brand audit attached and a trace of how it was built.

Brand Check closes the loop. Every output is scored against the brand for speed, volume and accuracy, so you can see the quality rather than take it on faith. And what it learns from the recurring conflicts and the near-misses feeds back into Brand Truth, so the source gets sharper the more work it sees.

So the whole system is a source written in markdown, skills that draw on it, output you can trace back to it, and a check that teaches it. The whole thing sits on a plain file server, with no build step.

What it can do, and who it's for.

Once Brand Truth exists, it produces brand-true work across almost any touchpoint: landing pages, decks, ads, emails, and brand checks that score an output and tell you exactly where it drifts. A website is just one thing it can make. The same source scales to the next one, and the next, at a fraction of the time, because nobody's re-explaining the brand from scratch.

It's built for teams and meant to be co-owned. Run it hosted with no install, or locally if your data rules need it. Brand, marketing, and engineering all work from one branch of one repo, and the same file serves all four: the Head of Brand sees the beliefs and the discipline, the product team sees how it turns into real work, the CTO sees how a request resolves, and the CMO can open any trace and audit it.

It won't replace the people who do the work, and it isn't trying to. It governs what a tool is allowed to produce and leaves the making to you. Craft is what trains Brand Truth, and Brand Truth is what scales the craft back out again. AI earns its place where decisions repeat and quality can be governed. It has no business on the last 10 percent, the call between good and almost-good, which stays stubbornly human.

Stop prompting and praying.

Left to itself, even the best model is a guess. You describe the brand from memory and hope what comes back is close. Put the Brand Context Protocol between the model and your tools, and the same model starts producing work that already knows the brand, inside the software your team uses anyway.

Why it's different.

We build it with you.

It's as much a way of working as a thing you buy. We build it together over about 14 weeks, in four phases. Along the way we teach your team to run it themselves, making real assets with the tools they already have, like Claude, on top of the protocol. Each phase stands on its own, and nothing moves to the next without evidence that the last one worked.

The brands that compound through this shift won't be the ones with the fanciest AI. They'll be the ones with the most disciplined judgment running through their AI, consistently, everywhere, at speed, with the brand never leaving their hands.

We're not building better brand tools. We're making brands usable, keeping them yours, and letting the system get smarter every time it's used. That's the Brand Context Protocol. Anyone can have the structure. The taste and the curation that make it worth anything are the part we bring.

Questions we get asked.