Copilot Studio Agent Patterns

A collection of production-ready patterns for building autonomous agents on the Microsoft Copilot Studio + Power Platform stack. Each solution is a self-contained implementation with prompts, schemas, provisioning scripts, and UI components.

Solutions

Solution	Description	Components
Intelligent Work Layer	Intelligent work layer that intercepts email, Teams, and calendar signals — triaging, researching, and preparing draft responses autonomously via a MARL pipeline with 18 agent prompts, learning from user behavior across 9 Dataverse tables, and rendering results on a PCF React dashboard with WCAG AA compliance. Design docs cover architecture, learning, and UX enhancements.	Copilot Studio agent (18 prompts), Power Automate flows, Dataverse (9 tables with learning system), PCF React dashboard, Canvas app, OneNote (optional)
Email Productivity Agent	Gmail-like follow-up nudges and smart snooze for Outlook — automatically reminds you to follow up on unreplied emails and unsnoozes threads when new replies arrive, with CLI regression harnesses for deterministic dry-run validation	Power Automate flows, Dataverse, Teams Adaptive Cards, optional Copilot Studio assets, HTTP harnesses
Agent Cost Governance — PAYGO	Leadership-quality PAYGO cost visibility for Copilot Studio agents — budget dashboards, alerts, and FSI regulatory alignment using Azure Cost Management + Power BI	Azure Cost Management, Power BI, PowerShell provisioning, ARM templates, FSI governance artifacts

Prerequisites

• PAC CLI — Power Platform CLI for provisioning and deployment
• Node.js 18+ — Required for Power Apps Component Framework (PCF) component builds
• PowerShell 7+ — Required for provisioning scripts
• Power Platform environment (Copilot Studio capacity is only required for solutions that actively invoke live agent steps)

Repository Structure

Each solution lives in its own top-level folder with a consistent structure:

<solution-name>/
├── docs/          # Setup guides and flow documentation
├── prompts/       # Agent system prompts
├── scripts/       # Provisioning and deployment PowerShell scripts
├── schemas/       # JSON schemas and table definitions
├── templates/     # HTML templates (e.g., OneNote page templates)
└── src/           # Code components, flow definitions, and topic YAMLs

Getting Started

1. Pick a solution from the table above
2. Read its README.md for architecture overview
3. Follow docs/deployment-guide.md for step-by-step setup

Contributing

Each solution should include:

• System prompts with clear input/output contracts
• JSON schemas for structured data
• Provisioning scripts for repeatable environment setup
• Documentation sufficient for a Power Platform maker to implement the flows