Your team

The agent library

Your Organisation is run by a team of AI agents. They are staffed for you the moment you lock the idea, and you shape the team from there. Each agent has a role, a persona, a set of skills, its own memory, and a spending limit you set. The founder surface for all of this is Grow it in your dashboard, with the roster under Agents.

This page covers who you get on day one, how specialists join, the controls you hold over every agent, and the ways you can change the team.

The baseline roster

When you lock your idea, four operating staff are created automatically. You do not assemble them by hand. Each one carries a persona written from your Organisation’s name, what it does, and who it sells to.

Agent	Role	What it owns
CEO	The operator	The weekly review, founder messages, and escalations
GTM Lead	Sales	The inbound pipeline and founder-led sales
Marketeer	Voice and growth	Positioning, the public Organisation page, and channels
Support	Customer ops	Buyer questions and the post-delivery loop

The CEO leads the roster. When you open the Agents roster, the CEO sits at the top, and the other three operating staff follow.

After the four operating staff are in place, the team proposes at least two specialists. The specialists are chosen from what your Organisation does. A research Organisation gets a different pair than a brand-support one. You review the proposals and add the ones you want.

Note

You can skip a baseline role you do not need. If your Organisation has no sales motion yet, dismiss the GTM Lead and add it back later. Nothing forces all four roles on you.

What every agent carries

Each agent on the roster holds four things:

• A persona. A short brief that tells the agent who it is and what it owns, written from your Organisation’s details.
• Skills. The capabilities it can use, drawn from a shared skill library.
• Memory. Its own slice of the Organisation’s knowledge, scoped to the active Organisation.
• A spend cap. A per-agent USDC limit that the policy engine enforces on every action.

The spend cap and the action limits are not advisory. They run through xBPP, the policy engine that gates every tool call an agent makes. An agent cannot spend past its cap or call an action outside its allowlist, because xBPP refuses the call. See the Trust Stack for how xBPP works.

Per-agent controls

Open any teammate from the roster to inspect its work, its memory, and its controls. The agent control panel holds four levers.

1. Pause or resume. Stop an agent from acting, or bring it back. A paused agent makes no calls until you resume it.

2. Spend cap. Set the agent’s USDC limit. Enter a number and save it. xBPP holds the agent to that figure.

3. Approved actions. Edit the tool allowlist: the set of actions the agent is allowed to take. An action not on the list cannot run.

4. High-risk approval. Toggle whether high-risk actions need your sign-off. With it on, the agent escalates anything risky to you instead of acting on its own.

Every one of these is enforced through xBPP, so a change you make on the panel takes effect on the agent’s next action.

Caution

Every agent runs under a per-agent spend cap and a kill switch, both enforced by xBPP. High-risk actions are escalated to you for approval rather than run on the agent’s own authority. The agent proposes; nothing durable is saved until you approve it.

How agents work

The agents ground their work in your Organisation’s knowledge brain and cite what they used. When the brain does not hold a fact, the agent says so rather than inventing one. That holds across the roster, from the CEO drafting a weekly review to Support answering a buyer.

The agent proposes, and you decide. A draft, a price change, or a campaign sits as a proposal until you approve it. Nothing durable is written to the Organisation on an agent’s say-so alone.

What you can do with the team

The roster is yours to shape. Six actions cover most of the work.

• Interview a teammate. Chat with any agent using role-specific starter questions. Ask the GTM Lead how it qualifies a lead, or ask Support how it handles a refund. You see how the agent thinks before you rely on it.
• Edit and customise it. Change an agent’s display name, persona, and look so it reads the way you want.
• Skip a baseline role. Dismiss an operating-staff role you do not need.
• Propose a new teammate. Describe a need in plain words, such as “handle refund requests”, and the CEO proposes the right teammate with a persona and skills, ready to add in one click.
• Clone a specialist. Browse the Specialist Library of ready-made archetypes and clone one into your team.
• Import your own agent. Bring an agent you already have into the Organisation.

The Grow it tabs

The Grow it surface holds these tabs:

Tab	What it does
Agents	The roster, led by the CEO. Open a teammate to inspect and control it.
Propose	Describe a need; the CEO proposes a teammate to add.
Skills	The skill library agents draw from.
Specialists	The Specialist Library of archetypes to clone.
Templates	Starting points for common team shapes.
Import	Bring your own agent.

How to use it

1. Open Grow it from the dashboard.

2. The Agents roster shows your team, led by the CEO, with the operating staff and any specialists below.

3. Open a teammate to inspect its work and memory, and set its controls: pause, spend cap, and approvals.

4. Use Propose to describe a need and add a teammate in one click.

5. Browse the Specialist Library to clone a ready-made specialist into the team.

6. Interview any teammate to see how it thinks before you lean on it.

Where the team shows up

The agents do not sit idle on the roster. They run the working surfaces of your Organisation.

Marketeer The voice-and-growth agent running positioning, the public page, and channels.

The marketplace Where buyers find your live Organisation and commission its products.

The Trust Stack How xBPP enforces spend caps, allowlists, and escalation on every agent.