Microsoft Copilot Studio: AI Agents, RAG, MCP & AI Security

Explore generative AI fundamentals and large language models, differentiate them from predictive AI, and learn tokens, chat completions API requests, multi-modality, and system and user props.

Explore the evolution from artificial intelligence to generative AI and large language models. See how transformer-based foundation models power Copilot and GPT, with fine-tuning enabling business use cases.

Learn core generative AI jargons, including tokens, system prompts, user prompts, chat completions API, and the difference between unimodal and multimodal models, with emphasis on token-based costs.

Define compound AI systems and AI agents that coordinate LLMs, planning, and tools to source data and perform tasks within Microsoft Copilot.

Explore Microsoft Copilot Studio and its agentic ai capabilities through licensing insights and hands-on labs, building embedded and standalone agents, and designing multi-agent workflows to extend Microsoft 365 Copilot.

Master Microsoft Copilot Studio, a no-to-low-code platform for citizen developers to create declarative and custom agents, with 1200+ data connectors and Microsoft Foundry integration.

Explore Copilot Studio licensing, including pay-as-you-go and a 25,000-credit monthly commitment, plus a free trial; credits cost 2 per answer and 5 per agent action.

Navigate Microsoft Copilot Studio, sign in with a licensed account, and create or publish embedded or standalone agents and flows, including natural-language workflows, rest APIs, and custom connectors.

Microsoft released a public preview of Copilot Studio's UI upgrade at Build 2026; use the classic UI for labs, as the new UI is not yet generally available.

Explore how the generative AI orchestrator and agentic planning drive runtime decisions in Copilot Studio, sequencing a SharePoint HR policies knowledge tool and an MSN weather data connector action plugin.

Create a web-grounded agent in Microsoft Copilot Studio by binding a public website (Udemy profile) as knowledge, then configure triggers, channels, and publish as a standalone agent.

Create a SharePoint embedded agent for Microsoft 365 Copilot. Link SharePoint sites and PDF documents, plus a Word document, as a knowledge base, then publish as embedded and standalone agents.

Compare embedded and standalone agents in Microsoft Copilot Studio, publish to Teams, and use embedded agents in the same Copilot chat with the @ symbol for contextual 365 app support.

Learn how to add the MSN weather data connector as an action to an agent in Copilot Studio, configure get current weather, and synthesize responses using generative AI.

Explore how prompt engineering sharpens prompts to maximize the AI's potential to deliver accurate and meaningful results in Copilot, focusing on goal, context, expectations, and source.

In this hands-on lab, learn to use a custom prompt action as a tool in Microsoft Copilot Studio to turn resumes into structured JSON outputs for HR screening.

Explore building autonomous agent flows in Copilot Studio by composing Power Automate flows with a resume screener custom prompt action, routing results via email.

Learn to embed AI Builder sentiment analysis in a Microsoft Copilot Studio agent flow, classifying reviews by sentiment with confidence scores and triggering emails or data updates.

Explore how model context protocol servers connect ai agents to enterprise knowledge and tools, offering a universal adapter that removes hard-coded rest calls and enables dynamic, interoperable access.

Shows building a Microsoft Learn MCP server agent in Copilot Studio with the model context protocol, leveraging Microsoft Learn modules for a three-month AI102 study plan.

Learn to create multi-agent workflows in Copilot Studio by orchestrating a parent agent with a MCP child agent to query the MLSLearn platform.

Explore the agent-to-agent protocol (A2A), how it enables interoperable communication between agents across frameworks, and how it complements MCP servers for open agent ecosystems.

Learn to set up inter-agent communication in Copilot Studio by connecting a client to a remote A2A server, defining the endpoint URL, and configuring description and authentication.

Learn to configure escalation topics to route to a human agent via Power Automate flows, and use count-based fallback topics to handle unknown intents in Copilot Studio.

Explore the analytics tab in Microsoft Copilot Studio to monitor a published agent’s performance, including conversation sessions, engagement, customer satisfaction, and usage metrics, plus savings and ROI insights.

Explore the Copilot Studio settings tab to configure generative AI options, agent details, security, channels, and data connections for a grounded web knowledge agent.

Extend Microsoft 365 Copilot with Copilot Studio or Azure AI Foundry to build domain-specific agents that solve department-level business problems and boost ROI.

Learn to build Prom Flows microservices in Azure AI Foundry with low-code prompts and Python, create a consumable API, and run it within a Copilot Studio agent flow.

Explore how Azure AI Foundry Studio Hub unifies model catalog, agents, and llm ops to build enterprise agents with diverse foundational models, orchestration, and serverless API hosting.

Deploy a hub-based foundry project to enable prompt flow within a centralized AI hub, then explore the Foundry project studio, model catalog, AI services, and PromFlow features.

Deploy chat completion models from a vendor-agnostic catalog in AI Foundry, configure global standard deployments, and test endpoints in the chat playground with code snippets.

Discover Microsoft PromFlow, a low-code Python-based microservice that prototypes and deploys AI applications with LLMs, emphasizing prompts, Python components, and API-driven collaboration.

Build a named entity recognition flow using a two-component prompt flow and a Python data cleanse step, deploy GPT-4.1 Kuljot, and verify location and job role extractions.

Deploy the NER PromFlow to a real-time managed endpoint on a standard D2ASV4, with primary and secondary keys for authenticated API requests and support for traffic management and ab testing.

Perform a hands-on lab to query the real-time PromFlow endpoint via rest api calls, with headers and an api key, to infer named-entity recognition for entity-type and user-query location Paris.

Join this hands-on lab to connect a Copilot Studio agent to a PromFlow endpoint, call the API, and perform named entity recognition via a Power Automate flow.

Explore retrieval augmented generation (RAC) to ground enterprise data in Copilot Studio, building vector indexes with Azure AI Search, and enabling multimodal RAC for textual and image data.

Explore retrieval augmented generation, or RAG, to ground LLMs in your enterprise data using vector embeddings, a retrieval pipeline, and augmented prompts, with Azure AI Search and multimodal RAG.

Learn to build multimodal RAG with Azure AI Search, ingesting PDFs and images, extracting text and image descriptions, generating vector embeddings, and preserving layout for citations in regulated documents.

Deploy a multimodal RAG pipeline on Azure by provisioning storage, enabling anonymous access for travel brochures, and configuring a search index with GPT-4.0 and AIDA-002 for a prompt flow chatbot.

Create a multi-modal rag ai search index by ingesting text and images from Azure Blob Storage and vectorizing them for search.

Connect a multimodal RAG index in Azure AI search with a Copilot Studio RAG agent to answer Maggie's travel queries through search results.

Explore Microsoft Foundry, a platform-as-a-service for building scalable pro-code agents and deploying models from Anthropic, Meta, and OpenAI. Integrate a Foundry agent with Copilot Studio in a cross-world, multi-agent workflow.

Microsoft Foundry unifies model catalog and agent service with Foundry IQ to build cloud-native or local AI agents, with governance, observability, and enterprise knowledge access.

Explore the differences between hub-based and standalone Microsoft Foundry projects, comparing organization-level collaboration, project-level isolation, and how Foundry SDK versions 1 and 2 support hub-based versus standalone setups.

Deploy a standalone Microsoft Foundry project in Azure portal to explore the new Foundry studio, agent service, and model catalog, including governance, monitoring, and cost projection.

Deploy your first LLM from the model catalog in Microsoft Foundry via a serverless API. Evaluate models by quality, safety, throughput, and cost on the leaderboard.

Create your demo agent in the Microsoft Foundry portal, enable versioning with saves, and configure prompts. Explore tracing, AI quality metrics, and attach tools like file search and memory store.

Explore Copilot Studio integration with the power platform and dataverse, outline learning outcomes, and capstone workflows from dataverse tables to agent-driven interactions and devops-ready solutions.

Explore Microsoft Power Platform, a no-code to low-code suite enabling canvas and model-driven apps, power pages, Power BI, and Power Automate for agentic AI with Copilot Studio and AI Builder.

Learn how Microsoft Dataverse serves as a secure cloud data store with standard and custom tables, RBAC, and the common data model, powering Dynamics 365, Power Apps, and Power Automate.

Navigate make.powerapps.com and the admin center to manage apps, flows, and Dataverse across environments; deploy solutions with devops pipelines and Copilot Studio while reviewing security and licensing.

Build a Dataverse custom table with Copilot Studio Agent and AI Builder sentiment analysis. Trigger agent flows to populate sentiment scores and assemble a managed solution for production deployment.

Create a Dataverse custom table named reviews with Copilot AI assistance, laying the foundation for a sentiment analysis AI builder flow triggered by a Copilot Studio agent.

Learn to build a Copilot Studio agent and flow that uses AI Builder sentiment analysis to populate a Dataverse review table with dynamic fields.

Learn about Microsoft Power Platform environments and solutions across the development to production lifecycle, including dev, sandbox, and region-aware strategies, with pipelines and managed versus unmanaged solutions.

Learn to build a managed Copilot Studio solution by bundling the agent, its flow, and the custom Dataverse table, then export as a managed solution and deploy via pipeline.

Explore data and ai security with Copilot Studio and Microsoft Purview, building information protection policies with Purview sensitivity labels, encryption, and access control for Microsoft 365 and Copilot agents.

Microsoft Purview provides the information protection and governance layer securing Copilot for Microsoft 365, enabling data discovery, classification, sensitivity labels, and DLP across multi-cloud environments.

Learn how to implement information protection with Microsoft Purview in a Copilot for M365 environment, applying sensitivity labels and access controls in SharePoint across finance and HR.

Integrate Microsoft Purview information protection and sensitivity labeling policies with Copilot Studio agents by applying labeling and access controls to SharePoint grounding knowledge.

Explore hands-on information protection in Microsoft Purview, creating sensitivity labels, auto-labeling policies, and classifiers to protect GST numbers and other sensitive data across SharePoint and M365.

Demonstrates integrating Purview information protection with a Copilot Studio agent that grounds a SharePoint Word document, respects the sensitivity label and shield indicator, and publishes to Teams.

Learn how DSPM for AI in Microsoft Purview helps you view policies, apply recommendations, and enable data discovery to secure AI workloads, Copilot for M365, and onboarded endpoints.

Discover how Copilot Studio and Microsoft Antra enable agent governance through zero trust principles, Agent ID, and conditional access policies to secure Microsoft 365 deployments at scale.

Explore how AI security and zero trust reinforce each other by applying pillars: verify explicitly, least privilege, and assume breach, with agent ID and Microsoft Anthra ID enabling governance.

Enable agent ID for all Copilot Studio agents in Power Platform Admin Center, then view Anthra agent ID in metadata to apply conditional access policies and privileged identity management policies.

Review agent id capabilities in the Antra admin center and map IDs to Copilot Studio agents. Note owners and sponsors, and explore global versus quarantined collections with Microsoft Graph API.

Create a conditional access policy in Microsoft Entry Admin Center to govern ai agents, selecting agent identities and attributes, and blocking access when an agent is at risk.

Enable threat detection for Copilot Studio in the Power Platform admin center to deploy Microsoft Defender XDR protection, guard against prompt injection, and control images and URLs in agent responses.