Four layers that turn Quick into a working agent. The mental model every later module hangs off.
Think of it like a kitchen: the Space is the kitchen, the Chat Agent is the recipe, a Flow or Schedule is the doorbell that triggers a run, and MCP is the oven — the thing that actually connects to real heat (real systems).
A container that holds documents and scopes what an agent is allowed to know. Your policy manuals, procedure PDFs, and reference data live here.
A saved, reusable prompt configuration — persona, goal, knowledge source, guardrails, and suggested prompts. The reusable unit of value (Module 5 goes deep here).
A wrapper that triggers a Chat Agent — on a schedule (e.g., every Monday 7am) or as a multi-step no-code workflow with branching logic.
The live tool connection to a real system — a core banking API, a document store, a market data feed — giving the agent something to actually act on.
| Layer | Finance Team Example |
|---|---|
| Space | "Compliance & Onboarding" Space holding the AML procedures, expense policy, and close procedure documents |
| Chat Agent | "Break Triage Agent" — classifies reconciliation breaks and drafts a disposition note |
| Flow / Schedule | A Flow that runs the Break Triage Agent every morning at 6am against the overnight reconciliation export |
| MCP Connector | A connector to the core banking system's transaction API, or to a public reference data source |
Every module after this one is really just "which layer are we in right now." Module 5 is entirely about layer 2. Module 6 covers layer 3 (Flows) plus Quick Research. The MCP deep-dive covers layer 4. Keeping this 4-layer picture in your head prevents over-building — most teams should max out layers 1 and 2 before ever reaching for a Flow or an MCP connector.