Your agents

An agent in WilsonOS is not a chat window. It is a long-running colleague with a name, a personality, a lane, and a boot contract — a defined set of things it knows the moment it wakes. There is a small team of them, and each has one job.

The team

Scout — Code Pineapple

The brand agent for a children’s book imprint. Playful, kid-book energy; knows the readers are parents of six-to-ten-year-olds. Owns everything Code Pineapple ships.

Prime — RiseBlox

The brand agent for a technology and education brand. Builder’s mindset — systems and frameworks. Owns everything RiseBlox ships.

Mia — across brands

The owner’s accountability coach and the steward of the system. Warm, direct. Keeps the plate honest, runs the weekly review, and can say what every brand is working on without loading every brand’s context.

The platform engineer

An engineering agent that builds the WilsonOS app itself. Its lane is the platform, not the brands; it never authors brand content.

Other brand agents exist in the running system and follow the same pattern: one brand, one agent, one voice.

Lanes

Each agent has a lane — the part of the system it is responsible for — and the lanes do not overlap:

Lane	Who	Delivers
Operations	Brand agents; Mia as cross-cutting lead	Business assets: books, campaigns, pages, videos — through skills and flows. Brand agents may also create and improve the skills and flows they use.
Platform	The engineering agent	The app, its data contracts, infrastructure, integrations.

A brand agent does not need to track what the platform team is shipping; when a capability lands, it appears in the agent’s toolkit. The platform agent does not decide what a brand says.

Identity

An agent’s identity is a small set of files: who it is and how it boots (its role, its brand, its startup checklist), its personality (how it talks), its schedule (the automations it owns), and a stable identifier that the rest of the system keys on. Renaming an agent keeps its history; deleting and recreating one mints a new identity that inherits nothing — history and permissions never transfer by accident.

Changing an agent’s identity or schedule is approval-gated: agents may propose changes to themselves, but the owner confirms them.

What an agent knows when it wakes

Every session — whether it starts from a chat message, a scheduled run, or an operator’s terminal — an agent boots with the same layered context, and nothing more:

1. Who it is and the house rules. Its identity and personality, and the workspace’s working discipline.
2. Who you are. Your profile: role, preferences, hard rules (“never X”, “always Y”). This layer outranks every other on conflict.
3. The brand essence. The one-page brand identity; deeper brand documents load on need.
4. Its plate and its memory. Open tasks assigned to it, feedback waiting on it, its most relevant facts, recent events, any automation that failed in the last two days, and a short narrative of what it was last doing.

That is promise 1 of the WilsonOS Five made concrete: knows you, your brands, and what’s on your plate today. Everything else — architecture references, full brand documents, the results ledger — is pulled in by the skill or flow that needs it, not preloaded. The boot is lean on purpose: it is paid on every conversation.

Two loaders, one bundle

Static identity loads as fixed documents. Live state — the plate, the memory digest, failed automations — is assembled by a program at boot, because it has to query the database and filter recent events, and it has to fire on every surface (chat, schedule, terminal). Deep detail: How agents run.

How you meet them

Through chat, mostly — from a laptop or the phone app. The agent replies in the conversation, shows its work as it runs a flow, and asks finite-choice questions with a real form when a decision is yours. Scheduled automations deliver their results to a conversation you chose. See Conversations.

live The roster, lanes, identity files, and the boot bundle are live and verified against the running system.

Related

Your brandsThe workspace each brand agent owns.

ConversationsHow you talk to an agent, and what it shows you while it works.

How agents runAlways-on processes, one runtime per conversation, isolation.

Session startThe boot contract in detail.