Your brands
Most agent tools organise around a conversation. WilsonOS organises around a brand. That one choice is what lets an agent stay in voice without prompting, keep two businesses from bleeding into each other, and start every session already knowing the business it serves.
What a brand holds
Section titled “What a brand holds”A brand is a workspace with its own brand stack — the documents that describe the business:
| Part | What it answers |
|---|---|
| Brand essence | The one-page identity: what it is, who it serves, how it sounds. Loads every session. |
| Products and offers | What the brand sells, how it is packaged and priced, where it is sold. |
| Customer | Who buys and why, in their own words — the phrases they use, the problems they name, what moves them from discovery to purchase. |
| Voice and visual rules | How the brand writes and looks. Applied to every customer-facing output automatically. |
| Strategy and positioning | The current bets: which channels, which audiences, which messages. |
| Lessons | What has been learned — from ratings, results, and the owner’s corrections — promoted deliberately into the brand’s shared knowledge. |
The brand stack is the Brand library of the agent’s memory: it outranks the agent’s general capabilities on any conflict. If a skill says “lead with the keyword” and the brand’s voice rules say “lead with the reader’s problem”, the brand wins.
One brand, one agent
Section titled “One brand, one agent”Every brand has exactly one agent as its operating officer. Scout runs Code Pineapple, a children’s book imprint. Prime runs RiseBlox, a technology and education brand. Each loads its brand’s essence at the start of every session and pulls deeper brand documents on need, following links from the essence, rather than loading everything up front.
This is a silo by design. Scout does not carry RiseBlox’s context, and Prime does not carry Code Pineapple’s. Awareness across brands — “what is everyone working on?” — is handled by the cross-cutting agent, Mia, without loading every brand’s stack.
Brand context makes the same method produce different work
Section titled “Brand context makes the same method produce different work”Skills and flows are brand-agnostic. A social-content flow is the same flow for every brand; what differs is the stack it loads. Running it as Scout produces playful, parent-facing copy in Code Pineapple’s palette; running it as Prime produces RiseBlox’s register. The method is shared and improved once; the output is always in brand. See Skills.
Setting up a brand
Section titled “Setting up a brand”- Describe it once. Products, audience, voice, visual rules, customer language, current strategy. The brand essence is the short version; the rest hangs off it.
- It gets an agent. The agent’s identity names the brand it owns; that binding is what scopes its sessions, its memory, and its access.
- Connect what it publishes through. Store listings, Pinterest, advertising accounts, email — each a named connection with its own scope (Connections).
- Decide who may see it. Access is granted on the brand, not on individual files (Access & permissions).
What does not belong to a brand
Section titled “What does not belong to a brand”- The owner’s own profile and preferences — how you like to work, your hard rules. That is the User library, shared across brands and outranking everything.
- Skills and flows — the shared catalogue, improved once for everyone.
- Private conversation memory — what you and an agent discussed stays scoped to you and that agent; it reaches the brand’s shared lessons only through deliberate promotion (What your agent remembers).
live Brand stacks, one agent per brand, essence-first loading, and brand-scoped access grants are live. direction Brand-level ownership on every record is being extended table by table ahead of multi-user use.
Related
Sources: AGENTS.md (brand silos), docs/architecture/memory.md (Brand library), docs/architecture/identity-access.md §3 · Last verified 2026-09-15