Brand Kit

Set your identity, logos, colours and tone of voice once in Settings → Brand. All 10 agents read the same brand kit — and none of them can change it.

Questa pagina non è ancora disponibile nella tua lingua — stiamo mostrando la versione in inglese.

Your brand kit is the one screen that upgrades all ten agents at once.

Fill it in once in Settings → Brand: who you are, how you sound, your logos, your colours. It lands in your Neural Core — the memory every agent reads on every run — so you never explain your business twice. And it's the one part of your memory the agents are structurally forbidden from editing.

What the brand kit is

The Neural Core is your business memory: contacts, products, files, rules, knowledge. Most of it grows on its own as you talk to your agents.

Your brand kit is different. It's the part you own outright:

You write it. Agents never extract brand facts from your conversations — that category is refused in code, not just discouraged in a prompt.
Every agent reads it. All 10 agents are mounted on the same Neural Core with no per-agent filter. There is no agent that has a different idea of your brand.
Agents never learn it on their own. Brand is the one category the background extractor is forbidden to touch, and the save tool refuses to create brand entries. Your brand changes when you change it in Settings → Brand.

Brand lives in Settings, not in the Neural Core grid

Contacts, Products and Files are managed at /brain. Your brand kit moved to Settings → Brand — it's a form, not a folder. That's the only place it can be edited.

What's in it

Five groups of field, saved as you type:

Section	Fields
Identity	Name, slogan, description, sector
Contact	Website, email, phone, address
Social	Instagram, Facebook, LinkedIn, X, TikTok
Visual	Light logo, dark logo, primary colour, secondary colour, accent colour, font
Voice	Tone, and a set of adjectives that describe how you sound

Nothing is mandatory. A new account starts with four brand entries scaffolded for you — Brand Identity, Visual Identity, Contact and Social Media — waiting to be filled.

Two logos, on purpose

Upload a light logo and a dark logo. When Ashley stamps your logo onto a post image she reaches for the light one first and falls back to the dark one. If you've uploaded neither, she asks you instead of inventing something.

Set up your brand kit

Open Settings and go to the Brand tab.
Start with Identity — your name, your slogan, one or two lines on what you do, your sector.
Fill in Voice: your tone, plus the adjectives you'd want a stranger to use about your writing. This is the field that changes the most output for the least typing.
Upload your light logo, then your dark logo.
Set your primary and secondary colours. (They default to a blue and a green until you change them.)
Add your contact details and social handles if you want agents to have them at hand.

There's no Save button. Every field auto-saves about half a second after you stop typing; logo uploads save immediately.

Voice is the highest-leverage field

"Direct, warm, no jargon, never use exclamation marks" costs you eight seconds and changes every email, caption, document and website your agents produce from that moment on.

How your agents use it

Two different things happen, and it's worth knowing which is which.

Your brand context is always there. Every agent loads your entire Neural Core — brand included — at the start of every run. Nobody has to look it up. Ask Lucy for a landing page and she already knows your sector and your slogan.

Your brand assets are fetched on request. When an agent needs the actual logo file or the actual hex codes, it calls up your brand kit and gets back your logo URLs, your primary and secondary colours, and — if it needs style direction — the images you've saved in your Neural Core knowledge.

That fetch is deliberately narrow. It fires when you explicitly ask for your logo or your colours. Asking for "something in our style" sends the agent to your saved images instead, which is usually what you actually meant.

Ask Ashley to stamp your logo on a post

Make me an Instagram post about our new autumn range, then put my logo on the image.

Ashley pulls the real logo file from your brand kit and overlays it on the generated image — she doesn't try to draw a logo, which is exactly how AI logos go wrong.

Ask Steve for a site in your colours

Build me a one-page site for my bakery. Use my brand colours and my slogan.

Steve can read your brand kit. He cannot write to it — his access to the Neural Core is read-only across the board.

Ask any agent what it thinks your brand is

What do you know about my brand?

Useful as a sanity check. Any of the ten will answer from the same source, so if one of them is wrong, they're all wrong, and the fix is one field in Settings → Brand.

Sell the boundary: brand is read-only to the AI

Most of your Neural Core is written by agents as you talk. Brand isn't — and that's enforced across every write path an agent has:

The save tool's schema doesn't even list brand as a category an agent could write to.
If an agent tries anyway, the save tool hard-rejects the write.
The update and delete tools refuse any entry whose category is brand, so an agent can't edit or remove a brand entry even by passing its id directly.
The background extractor — the thing that learns facts from your chats — is told to ignore brand facts entirely, and has a code guard that throws them out if it tries.

Nothing an agent does — in conversation, or by calling a tool directly — can rewrite your brand kit behind your back. It changes when you change it in Settings → Brand, and only then.

Which also means: nothing fills it in for you

Mentioning your tagline in a chat will not update your brand kit. That's the trade — total control, zero autofill. Set it once, deliberately.

What the brand kit does not do

Worth being straight about:

It doesn't apply your font. Your font and accent colour are stored on your brand profile, but the assets served to agents for images are your logos and your primary and secondary colours.
Social handles here are not connections. Typing your Instagram handle into the brand kit is reference information. To actually publish, connect the account from Integrations in the sidebar. (TikTok is a field here, but is not yet a connectable platform.)
It's one brand per account. There's no brand switcher for agencies running several clients.
Visual references have a ceiling. When an agent looks for style references it scans up to 50 of your saved knowledge images. If you have hundreds, name them well — see the tip below.

Tips

Name your images like you'd search for them. Style references are matched on the words in the name and description, so "packaging-flatlay-kraft" beats "IMG_4471".
Lock what's sensitive. Anything in your Neural Core can be marked Private, which puts a hard "do not mention this in any generated content" instruction in front of it for all ten agents at once. Useful for internal-only positioning notes.
Your slogan is not your description. Slogan is the line you'd put on a billboard; description is what you'd tell someone at a party. Agents use them for different jobs — fill in both.
Update it when you rebrand, not before. There's no version history on the brand kit; only the current state is stored.

Do all 10 agents really use the same brand kit?

Yes, and not by convention — by architecture. Your Neural Core is loaded for every agent on every run, with no per-agent filter anywhere in the code. There is no agent-specific copy of your brand to drift out of sync.

Can an agent change my brand?

Not on its own. Nothing an agent learns from your chats can reach your brand kit: the save tool's schema omits the brand category, the save tool's code rejects it, and the background fact-extractor has both a prompt rule and a code guard against extracting brand facts. Steve additionally has no write access to your Neural Core at all. Edit your brand in Settings → Brand.

Where do I save it?

You don't. Fields auto-save roughly half a second after you stop typing, and logo uploads save the moment they finish. Close the modal whenever you like.

Why does my logo appear on some images and not others?

Because it only appears when you ask. Your logo is applied when you explicitly request it — "put my logo on it", "use my brand colours". An agent won't stamp your logo onto every visual by default.

Does filling in my brand kit cost credits?

No. Reading and writing your Neural Core — brand kit included — costs zero credits. You're only billed for the work an agent actually does, at real cost. Memory is not a metered upsell.

I only have one logo. Is that a problem?

No. Upload it as the light logo. Agents try the light logo first and fall back to the dark one, so a single light logo always works. Add a dark version later if your logo disappears on dark backgrounds.

Troubleshooting

An agent says it can't find my logo. It's asking because both logo fields are empty — agents are told to ask rather than guess. Upload at least a light logo in Settings → Brand.

I told an agent my new slogan and it didn't stick. Expected. Brand is the one category agents never learn from conversation. Type it into Settings → Brand.

An agent used the wrong colour. Check your primary and secondary colours in Settings → Brand — until you set them, they hold default values.

An agent picked the wrong style reference. Style references are matched on the name and description of your saved images. Rename the image you wanted to something that matches how you'd ask for it, or point the agent at it by name.

Next: Your Neural Core · Files & knowledge · Integrations