Workers

Workers are task-oriented background agents that you can customize with custom prompts and tools, and—for Planner and Analyst—schedule. They give you building blocks for automation, reporting, and optimization that you can combine with the Success Agent and Playbooks to build end-to-end flows.

What you can build

Background automation — Run recurring or on-demand tasks (syncs, updates, notifications) with the Executor, or schedule the Planner and Analyst for reports and plans.
Reporting and analytics pipelines — Have the Analyst produce reports on store data; feed those into the Planner to generate improvement plans.
Optimization loops — Planner suggests opportunities; Executor applies changes; Playbooks can define steps that direct Workers (e.g. Executor) to carry out those changes when the merchant runs the playbook. You or the Success Agent can review or advise.
Custom workflows — Give Workers MCP Tools and custom prompts so they integrate with your APIs, internal tools, or third-party services.

Workers share the same MCP Tools and Guidelines as the Success Agent, so one integration or rule set applies across chat and automation.

Executor

The Executor implements actions. As a partner or developer you can:

Assign write tools — Grant the Executor access to specific tools (e.g. update products, manage campaigns) so it can apply changes safely and within the scope you define.
Carry out playbook steps — Playbooks consist of steps that assign tasks to Workers; each step can be assigned to a different Worker (Executor, Planner, Analyst). Steps that assign work to the Executor might have it apply changes or call tools. When the merchant runs a playbook, the relevant Worker performs each step in order.
Combine with Planner — Use the Planner to produce a list of tasks; have the Executor implement them (optionally with merchant or Success Agent approval).

The Executor is the “do” layer: it performs the concrete tasks. You control what it can do by choosing which tools it has access to and which playbook steps can assign work to it.

Planner

The Planner produces tasks and improvement ideas. You can:

Schedule it — Run the Planner on a schedule (e.g. weekly) to generate suggestions, opportunities, or to-do lists based on store data and Guidelines.
Feed it Analyst output — Point the Planner at Analyst reports so it creates plans from real data (e.g. “given these metrics, suggest campaign or inventory actions”).
Output tasks for Executors or playbooks — The tasks the Planner produces can be handed to the Executor or turned into playbook steps so the full loop is: Analyst → Planner → Executor (or playbook steps that direct the Executor).

The Planner is the “decide what to do” layer. It doesn’t perform actions itself; it proposes them. That makes it safe to run often and to expose to the Success Agent (e.g. “What should I focus on this week?”).

Analyst

The Analyst analyzes store data and analytics. You can:

Schedule reports — Run the Analyst on a schedule to produce regular reports (performance, inventory, orders, etc.).
Feed Planners — Use Analyst reports as input to the Planner so improvement plans are grounded in actual data.
Customize with prompts and tools — Give the Analyst custom prompts and MCP Tools so it can pull in external data or format reports the way your product or merchants need.

The Analyst is the “observe and summarize” layer. It produces reports; other Workers or the Success Agent can act on those reports.

Custom prompts and tools

All Workers support:

Custom prompts — Shape what they optimize for, how they phrase output, and what they focus on. Combined with Guidelines, you get consistent, on-brand behavior.
Tools — Workers use the same MCP Tools as the Success Agent. Connect your APIs, CRMs, or internal systems once; both chat and background automation can use them.

Scheduling (where supported) lets you run Planner and Analyst on a cadence so merchants or your app always have fresh suggestions and reports.

Combining with the rest of AI Commerce

Success Agent — Can call Workers (e.g. “Run the Analyst report”, “What does the Planner suggest?”). You build one chat surface that orchestrates all three.
Playbooks — Playbooks define steps that assign tasks to Workers; each step can target a different Worker. When the merchant runs a playbook, those steps are carried out in order. You get predictable, repeatable flows that use the same MCP and guidelines.
Guidelines — Apply to all Workers so automation stays within your rules and voice.

Overview: AI Commerce → | Success Agent → | Playbooks → | MCP Tools → | Guidelines →