The team is not a contact list next to the chat. It is what makes the chat useful: identities, skills, rules and memory that persist across turns and stores.Six identitiesShared memory10 skillsLevels 1–3
The numbers
The team as shipped
Named identities
6
Native skills
10
Fact scopes
3
Public profiles
/agents
Definition
What the team is
@Atlas orchestrates. @Maya, @Marco, @Otis, @Faye and @Sam each own a domain. They share OrgFact, StoreFact and UserFact memory extracted from every conversation. Skills are installable playbooks available to every agent. Rules live on the store and the organisation. Autonomy is a level you set per agent, from propose-only to execute-inside-scope.
Atlas, Maya, Marco, Otis, Faye and Sam share OrgFact, StoreFact and UserFact. Custom agent editing in the UI is still next-release copy, not a Features claim.
Roster
Six specialists, one memory
Full profiles live on /agents. This page is the operating model, not a second roster.
@AtlasOrchestrator. Routes work and holds the approval default.
Five specialistsMaya, Marco, Otis, Faye, Sam. Domain roles, not decorative avatars.
SkillsTen native skill.md packages. System manifest skill ids are display until wired.
MemoryScoped facts plus MemoryEvent audit. No Episode table.
The mechanism
How it compounds
Each turn teaches the next. Each skill multiplies every agent.
IdentitiesSix public profiles. Delegation tools hand work between specialists when team mode is on. Public pages live under /agents.
MemoryAfter each chat turn, facts are extracted and stored. The next prompt retrieves them. The same memory feeds WhatsApp and the dashboard.
SkillsTen filesystem skills ship with the platform. Install one and every agent on every store can use it. Custom skills sit beside them.
AutonomyLevels 1 to 3. Propose, approved writes, autonomous inside scope. A fourth level that runs the business unattended is on the roadmap, not a setting.
Why a team
When six beats one
HandoffsMerchandising and ops stay specialised without losing shared context.
Rules and skillsPlaybooks attach to identities instead of one mega-prompt.
The right case
When to use it
Use it whenWhen you want the OS to remember your brand and keep working with the same standards across agents, not start from zero every morning.
Do not use it whenWhen you have one question and no store connected yet. Use the home chat. The team compounds once there is something to remember.
Autonomy
Levels that exist
1, 2 and 3 onlyThe runtime type is 1|2|3. Anything that looks like a fourth setting is wrong.
The limits
Limitations
What the feature does not do. Knowing beforehand costs less than finding out after.
Autonomy stops at level 3. There is no operator setting for a fully unattended business.
Platform safety rules are not toggleable in the UI. Store and org rules are.
Skills on System manifests are display-only. They do not auto-wire tools or billing.
Memory extraction is best-effort. Failures are swallowed so a turn still completes.
Fenced
Not yet
Custom agent editor as a first-class Features claim.
On /agents, one page each. This page is the operating model. Those pages are the people.
Where did the Skills page go?AI Team
Here. A skill without a team that uses it is a file on disk. The team page is where it belongs.
Is this a separate knowledge product?AI Team
No. Knowledge indexing and memory facts are the brain of the team. Splitting them into their own menu rows made the OS look like a toolbox.
Can an agent publish to my live theme alone?AI Team
Only at the autonomy level you set, and critical ops stay gated. Default is level 1: it proposes, you approve.
@Atlas audits the catalog the day the store is connected and ships a CRO plan, instead of handing back a list of things to go read. Audit to plan, same session.
BoostEcom DemoDemo
Founder
One operator, the full stack: Maya runs the email calendar, Marco curates collections, Otis watches fulfilment. One team, one thread.
BoostEcom DemoDemo
Head of Growth
The work gets done rather than suggested. Voice mode with @Atlas turns the morning check-in into a conversation. It operates.
BoostEcom DemoDemo
Store owner
@Maya rewrites product descriptions, queues them for review, pushes the approved batch live, then runs the campaign around them. End-to-end, with approval gates.
BoostEcom DemoDemo
E-commerce lead
Every action the AI Team takes is graded against real Shopify Analytics after the observation window closes. Outcome attribution, not vibes.
BoostEcom DemoDemo
Marketing lead
OAuth scopes, audit log, store-scoped credits, no shadow data. The agents work inside the admin, not around it.
BoostEcom DemoDemo
Engineer
Permission mode decides what ships on its own and what waits for a human. Autonomy you dial in.
BoostEcom DemoDemo
Operator
Multi-store from one console, each with its own connectors, memory and budget. Per-store context.
BoostEcom DemoDemo
Agency owner
@Atlas audits the catalog the day the store is connected and ships a CRO plan, instead of handing back a list of things to go read. Audit to plan, same session.
BoostEcom DemoDemo
Founder
One operator, the full stack: Maya runs the email calendar, Marco curates collections, Otis watches fulfilment. One team, one thread.
BoostEcom DemoDemo
Head of Growth
The work gets done rather than suggested. Voice mode with @Atlas turns the morning check-in into a conversation. It operates.
BoostEcom DemoDemo
Store owner
@Maya rewrites product descriptions, queues them for review, pushes the approved batch live, then runs the campaign around them. End-to-end, with approval gates.
BoostEcom DemoDemo
E-commerce lead
Every action the AI Team takes is graded against real Shopify Analytics after the observation window closes. Outcome attribution, not vibes.
BoostEcom DemoDemo
Marketing lead
OAuth scopes, audit log, store-scoped credits, no shadow data. The agents work inside the admin, not around it.
BoostEcom DemoDemo
Engineer
Permission mode decides what ships on its own and what waits for a human. Autonomy you dial in.
BoostEcom DemoDemo
Operator
Multi-store from one console, each with its own connectors, memory and budget. Per-store context.
BoostEcom DemoDemo
Agency owner
The work gets done rather than suggested. Voice mode with @Atlas turns the morning check-in into a conversation. It operates.
BoostEcom DemoDemo
Store owner
@Maya rewrites product descriptions, queues them for review, pushes the approved batch live, then runs the campaign around them. End-to-end, with approval gates.
BoostEcom DemoDemo
E-commerce lead
Every action the AI Team takes is graded against real Shopify Analytics after the observation window closes. Outcome attribution, not vibes.
BoostEcom DemoDemo
Marketing lead
OAuth scopes, audit log, store-scoped credits, no shadow data. The agents work inside the admin, not around it.
BoostEcom DemoDemo
Engineer
Permission mode decides what ships on its own and what waits for a human. Autonomy you dial in.
BoostEcom DemoDemo
Operator
Multi-store from one console, each with its own connectors, memory and budget. Per-store context.
BoostEcom DemoDemo
Agency owner
@Atlas audits the catalog the day the store is connected and ships a CRO plan, instead of handing back a list of things to go read. Audit to plan, same session.
BoostEcom DemoDemo
Founder
One operator, the full stack: Maya runs the email calendar, Marco curates collections, Otis watches fulfilment. One team, one thread.
BoostEcom DemoDemo
Head of Growth
The work gets done rather than suggested. Voice mode with @Atlas turns the morning check-in into a conversation. It operates.
BoostEcom DemoDemo
Store owner
@Maya rewrites product descriptions, queues them for review, pushes the approved batch live, then runs the campaign around them. End-to-end, with approval gates.
BoostEcom DemoDemo
E-commerce lead
Every action the AI Team takes is graded against real Shopify Analytics after the observation window closes. Outcome attribution, not vibes.
BoostEcom DemoDemo
Marketing lead
OAuth scopes, audit log, store-scoped credits, no shadow data. The agents work inside the admin, not around it.
BoostEcom DemoDemo
Engineer
Permission mode decides what ships on its own and what waits for a human. Autonomy you dial in.
BoostEcom DemoDemo
Operator
Multi-store from one console, each with its own connectors, memory and budget. Per-store context.
BoostEcom DemoDemo
Agency owner
@Atlas audits the catalog the day the store is connected and ships a CRO plan, instead of handing back a list of things to go read. Audit to plan, same session.
BoostEcom DemoDemo
Founder
One operator, the full stack: Maya runs the email calendar, Marco curates collections, Otis watches fulfilment. One team, one thread.
BoostEcom DemoDemo
Head of Growth
Every action the AI Team takes is graded against real Shopify Analytics after the observation window closes. Outcome attribution, not vibes.
BoostEcom DemoDemo
Marketing lead
OAuth scopes, audit log, store-scoped credits, no shadow data. The agents work inside the admin, not around it.
BoostEcom DemoDemo
Engineer
Permission mode decides what ships on its own and what waits for a human. Autonomy you dial in.
BoostEcom DemoDemo
Operator
Multi-store from one console, each with its own connectors, memory and budget. Per-store context.
BoostEcom DemoDemo
Agency owner
@Atlas audits the catalog the day the store is connected and ships a CRO plan, instead of handing back a list of things to go read. Audit to plan, same session.
BoostEcom DemoDemo
Founder
One operator, the full stack: Maya runs the email calendar, Marco curates collections, Otis watches fulfilment. One team, one thread.
BoostEcom DemoDemo
Head of Growth
The work gets done rather than suggested. Voice mode with @Atlas turns the morning check-in into a conversation. It operates.
BoostEcom DemoDemo
Store owner
@Maya rewrites product descriptions, queues them for review, pushes the approved batch live, then runs the campaign around them. End-to-end, with approval gates.
BoostEcom DemoDemo
E-commerce lead
Every action the AI Team takes is graded against real Shopify Analytics after the observation window closes. Outcome attribution, not vibes.
BoostEcom DemoDemo
Marketing lead
OAuth scopes, audit log, store-scoped credits, no shadow data. The agents work inside the admin, not around it.
BoostEcom DemoDemo
Engineer
Permission mode decides what ships on its own and what waits for a human. Autonomy you dial in.
BoostEcom DemoDemo
Operator
Multi-store from one console, each with its own connectors, memory and budget. Per-store context.
BoostEcom DemoDemo
Agency owner
@Atlas audits the catalog the day the store is connected and ships a CRO plan, instead of handing back a list of things to go read. Audit to plan, same session.
BoostEcom DemoDemo
Founder
One operator, the full stack: Maya runs the email calendar, Marco curates collections, Otis watches fulfilment. One team, one thread.
BoostEcom DemoDemo
Head of Growth
The work gets done rather than suggested. Voice mode with @Atlas turns the morning check-in into a conversation. It operates.
BoostEcom DemoDemo
Store owner
@Maya rewrites product descriptions, queues them for review, pushes the approved batch live, then runs the campaign around them. End-to-end, with approval gates.
BoostEcom DemoDemo
E-commerce lead
Permission mode decides what ships on its own and what waits for a human. Autonomy you dial in.
BoostEcom DemoDemo
Operator
Multi-store from one console, each with its own connectors, memory and budget. Per-store context.
BoostEcom DemoDemo
Agency owner
@Atlas audits the catalog the day the store is connected and ships a CRO plan, instead of handing back a list of things to go read. Audit to plan, same session.
BoostEcom DemoDemo
Founder
One operator, the full stack: Maya runs the email calendar, Marco curates collections, Otis watches fulfilment. One team, one thread.
BoostEcom DemoDemo
Head of Growth
The work gets done rather than suggested. Voice mode with @Atlas turns the morning check-in into a conversation. It operates.
BoostEcom DemoDemo
Store owner
@Maya rewrites product descriptions, queues them for review, pushes the approved batch live, then runs the campaign around them. End-to-end, with approval gates.
BoostEcom DemoDemo
E-commerce lead
Every action the AI Team takes is graded against real Shopify Analytics after the observation window closes. Outcome attribution, not vibes.
BoostEcom DemoDemo
Marketing lead
OAuth scopes, audit log, store-scoped credits, no shadow data. The agents work inside the admin, not around it.
BoostEcom DemoDemo
Engineer
Permission mode decides what ships on its own and what waits for a human. Autonomy you dial in.
BoostEcom DemoDemo
Operator
Multi-store from one console, each with its own connectors, memory and budget. Per-store context.
BoostEcom DemoDemo
Agency owner
@Atlas audits the catalog the day the store is connected and ships a CRO plan, instead of handing back a list of things to go read. Audit to plan, same session.
BoostEcom DemoDemo
Founder
One operator, the full stack: Maya runs the email calendar, Marco curates collections, Otis watches fulfilment. One team, one thread.
BoostEcom DemoDemo
Head of Growth
The work gets done rather than suggested. Voice mode with @Atlas turns the morning check-in into a conversation. It operates.
BoostEcom DemoDemo
Store owner
@Maya rewrites product descriptions, queues them for review, pushes the approved batch live, then runs the campaign around them. End-to-end, with approval gates.
BoostEcom DemoDemo
E-commerce lead
Every action the AI Team takes is graded against real Shopify Analytics after the observation window closes. Outcome attribution, not vibes.
BoostEcom DemoDemo
Marketing lead
OAuth scopes, audit log, store-scoped credits, no shadow data. The agents work inside the admin, not around it.
