Skip to content

Glossary

Key terms used throughout this book.

Term Definition
A2A (Agent2Agent) Open protocol for inter-agent communication. Copilot Studio supports connecting to A2A agents (preview).
Adaptive Card A JSON-based UI card format for rendering interactive content (forms, buttons, images) in chat channels. Schema 1.5 for Teams, 1.6 for Web Chat.
Agent Instructions The system prompt (Markdown text in GptComponentMetadata.instructions) that defines the agent's personality, capabilities, and behavioral rules.
Azure AI Search Microsoft's enterprise search service supporting full-text, vector, and hybrid search. Used as an advanced knowledge source in Copilot Studio (GA since May 2025).
Child Agent A specialist agent connected to and invoked by a main (orchestrator) agent. Routes based on name, description, and orchestrator instructions.
ConversationStart A topic trigger (OnConversationStart) that fires when a new conversation begins. Does not fire in M365 Copilot.
Custom Connector A Power Platform connector created from an OpenAPI specification. Reusable across agents, flows, and Power Apps.
Dataverse Microsoft's enterprise data platform (part of Power Platform). Used for state persistence, configuration, and business data storage.
Direct Line Bot Framework API for programmatic communication with a bot/agent. Used for custom Web Chat implementations and automated testing.
Environment Variable A Power Platform configuration value (Env.Name) that can differ between environments (dev/test/prod). Used for URLs, API keys, and tenant-specific settings.
Fabric Data Agent A Microsoft Fabric agent that translates natural language into SQL/DAX/KQL queries against warehouses, lakehouses, and semantic models. Consumable as a connected agent in Copilot Studio (preview).
Generative Orchestration The AI-driven mode where the LLM automatically selects topics, tools, and agents based on user intent and available descriptions. Required for MCP tools.
Global Variable A conversation-scoped variable (Global.Name) that persists across topic switches within one conversation. Resets between conversations.
Integrated Vectorization An Azure AI Search feature that automatically chunks, embeds, and indexes content using a specified embedding model. Recommended for Copilot Studio connections.
KQL (Kusto Query Language) Microsoft's query language for Azure Data Explorer, Application Insights, and Fabric KQL databases.
LogCustomTelemetryEvent A Copilot Studio node that sends custom events to Application Insights for monitoring and analytics.
MCP (Model Context Protocol) An open standard for connecting AI agents to external tools and data sources. Copilot Studio supports MCP servers (GA) with prebuilt connectors for Dataverse, Outlook, GitHub, etc.
OnError A topic trigger that fires when an error occurs during conversation processing. Used for standardized error handling.
OnToolSelected A topic trigger used for child agent/tool definitions. Fires when the orchestrator routes to this agent.
OData A REST protocol for querying and manipulating data. Used in Dataverse and Azure AI Search APIs.
Power Fx The expression language used in Copilot Studio for conditions, calculations, and variable manipulation. Similar to Excel formulas.
Prompt Tool A Copilot Studio tool that executes an LLM prompt with defined inputs and outputs. Used for classification, summarization, and dynamic instruction generation.
RLS (Row-Level Security) A data security mechanism that restricts which data rows a user can see, based on their identity. Applied in Dataverse and Power BI.
Semantic Ranker An Azure AI Search feature that re-ranks initial search results using a deep learning model. Improves precision. Requires Standard tier.
SearchAndSummarizeContent A Copilot Studio node that searches knowledge sources and generates a grounded answer using an LLM. The primary generative answers mechanism.
SendActivity A Copilot Studio node that displays a message (text, card, or attachment) to the user.
Topic A discrete conversation flow in Copilot Studio, triggered by user intent (trigger phrases), events, or system conditions.
Trigger Phrases Natural language examples that train the agent's NLU model to recognize when a specific topic should activate.
VS Code Extension The ms-CopilotStudio.vscode-copilotstudio extension for editing agent YAML files locally with IntelliSense, then syncing to Copilot Studio.