Success Agent

The Success Agent is an intelligent assistant that merchants can chat with in the Finqu admin. As a partner or developer, you can build on it to deliver support tools, guided workflows, and smarter admin experiences—without building a chat UI from scratch.

What you can build

Merchant support and onboarding — Answer questions, walk through setup, and suggest next steps based on store state.
Guided workflows — Lead merchants through multi-step tasks (e.g. campaigns, inventory, reporting) with natural language and structured steps.
Custom admin experiences — Embed or extend the agent in your app so it feels like a single, intelligent interface for your merchants.
White-label support — Offer a branded assistant that uses your guidelines and skills so every merchant gets consistent, on-brand help.

The agent is a deep agent that can orchestrate predefined subagents and call Workers. You control its behavior through modes, custom skills, and guidelines.

Modes: ask, agent, and planning

The Success Agent supports three modes that you can use (or expose) depending on what you’re building:

Ask — Best for quick answers and lookups. The agent reads context and responds without taking actions. Use this when you want low-risk, informational interactions.
Agent — The agent can use tools, call workers, and perform actions (within the permissions you grant). Use this for support flows that need to change data or trigger workflows.
Planning — The agent breaks a request into a plan (steps, dependencies) and can then execute or present the plan for approval. Use this for complex, multi-step workflows where transparency and control matter.

When building support tools or guided flows, you can design experiences around these modes—for example, “ask first” for new users and “agent” or “planning” for power users who need automation.

Subagents and Workers

The Success Agent can delegate to:

Predefined subagents — Specialized agents for specific domains (e.g. products, orders, analytics). You can route merchant questions to the right subagent for better accuracy and control.
Merchant Workers — Executor, Planner, and Analyst workers that you or the merchant have configured. The agent can ask the Planner for improvement ideas, request reports from the Analyst, or ask the Executor to apply changes. The Success Agent can create Playbooks (e.g. help define or save a workflow) but cannot execute them; only the merchant can run a playbook.

This lets you build a single chat surface that feels like one assistant but is actually composed of specialized capabilities—all driven by your configuration and guidelines.

Custom skills (saved prompts)

Custom skills are predefined or saved prompts that extend what the agent can do. As a partner, you can:

Create reusable skills — e.g. “Review low-stock products”, “Summarize last week’s orders”, “Suggest campaign ideas”. Merchants (or your app) can invoke these by name or through natural language.
Ship skills with your solution — Preconfigure skills that match your app’s workflows so merchants get value immediately.
Let merchants create their own — Save one-off prompts as skills so repeated tasks become one-click or one-utterance operations.

Skills make the agent more predictable and easier to integrate into your product: you define the capability once, and it’s available in chat and (where supported) via API or workers.

Combining with the rest of AI Commerce

Use Guidelines to set tone, scope, and guardrails so the Success Agent stays on-brand and within safe boundaries.
Connect MCP Tools so the agent can call your APIs, internal tools, or third-party services.
Trigger or reference Workers (e.g. “Run the weekly Analyst report”) and create or refine Playbooks from a single conversation. The agent cannot run playbooks; only the merchant can run them.

Overview: AI Commerce → | Guidelines → | MCP Tools → | Workers →