Skip to content

Microsoft Copilot Studio

Microsoft Copilot Studio is a low-code SaaS studio for building agents and AI-driven workflows, connecting them to company data and systems, and publishing them to Teams, Microsoft Copilot, websites and other channels. Usage is billed in Copilot Credits, with some uses covered by a Microsoft 365 Copilot license.1

Also known as Copilot Studio, Power Virtual Agents (former product, now part of Copilot Studio)

Intended for
Business makers and citizen developers, Professional makers and Power Platform developers, IT and Power Platform administrators, Customer service teams
Runs on
Microsoft cloud service (SaaS), Commercial tenants, US Government Community Cloud (GCC)2

At a glance

What is it?
Copilot Studio is Microsoft's graphical, low-code tool for creating and managing AI agents, workflows and agent flows in one place, available as a web app at copilotstudio.microsoft.com. What was formerly Microsoft Power Virtual Agents is now included in Copilot Studio. Each agent or workflow runs on a harness that sets how it reasons and how it is billed: the GitHub Copilot harness for reasoning-heavy multi-step work, the Customer Conversation harness for customer service over voice and chat, and the standard harness for topic-based conversations. Building apps in Copilot Studio is in preview.1
What does it do?
Makers describe an agent in plain language, add instructions, knowledge sources and tools such as prebuilt or custom connectors and MCP servers, test it, and publish it. Workflows built in a drag-and-drop designer combine agents, tools and logic with human-in-the-loop steps; agent flows offer a Power Automate style format that can run alone or as an agent tool. Some agents can get their own account to work proactively. After publishing, teams use analytics, evaluations with test sets, and an agent inventory with role-based access and cost management.1
Who needs it?
Business units and IT teams that want to automate a process or answer questions with an agent without writing a full application, and that work mainly inside Microsoft 365 and Power Platform. Microsoft's guidance recommends Copilot Studio for scenarios such as an HR onboarding assistant over existing Microsoft 365 content. Customer service teams can use the Customer Conversation harness for voice and chat agents with escalation to humans.3
What does it need?12
  • A Copilot Studio license path: a tenant prepaid Copilot Credit pack plus free user licenses, pay-as-you-go through an Azure subscription billing policy, a Copilot Credit prepurchase plan, or a Microso
  • A Power Platform environment and admin access to the Power Platform admin center for billing policies and channel settings
  • Microsoft Entra ID for maker and user authentication
  • A modern browser: Chrome 91, Firefox 89 or Safari 16.4 or later
  • Note that a trial license allows building and testing but not publishing
What it is not
Copilot Studio is not a pro-code SDK: when developers need full control over models, orchestration or hosting, Microsoft points them to the Microsoft 365 Agents SDK or Microsoft Foundry. It is not Agent Builder, the simpler no-code builder inside Microsoft Copilot. It is not free: beyond some zero-rated uses for Microsoft 365 Copilot licensed users, agents consume Copilot Credits. Agents built with Copilot Studio cannot be published to the Microsoft Commercial Store through the Agents Toolkit, so it is aimed at internal and customer-facing use rather than selling agents to other companies.23

Availability and licensing. Generally available; the app creation experience and building with natural language are in preview (checked 26 Oct 2026). Usage is measured in Copilot Credits, which replaced messages on 1 Sep 2025. Microsoft's product page, checked 26 Oct 2026, lists tenant packs of 25,000 Copilot Credits at $200.00 per pack per month, plus pay-as-you-go and a prepurchase plan. A Microsoft 365 Copilot license covers building agents on the GitHub Copilot harness, and some agent use by licensed users is zero-rated.24

The problem it solves

Many business processes involve answering the same questions or moving information between systems, and building a custom application for each is slow. Earlier chatbot tools required scripting every conversation path.

Copilot Studio lets makers combine generative reasoning with structured automation in one low-code studio, connect to more than a thousand prebuilt connectors and MCP servers, and publish where users already work. Governance runs through the Power Platform admin center, Entra ID and Purview, so IT can control channels, environments and spending.4

How it works

Microsoft documents these building blocks:

  • Agents: follow instructions, draw on connected knowledge and use tools to act, deciding the next step from context.
  • Workflows: drag-and-drop automations where each step can reason and act, with built-in testing and human review.
  • Agent flows: the established flow format, similar to Power Automate, usable alone or as an agent tool.
  • Harnesses: the engine underneath. The GitHub Copilot harness reasons from instructions and picks tools; the Customer Conversation harness uses a primary agent with specialized sub-agents and can hand off to a human; the standard harness matches requests to designed topics and falls back to generative answers.
  • Channels: Teams and Microsoft Copilot, SharePoint, WhatsApp, custom websites, mobile apps, Facebook and Azure Bot Service channels such as Slack and Telegram.

Agents use Microsoft Entra ID authentication by default for Teams, Power Apps and Microsoft Copilot. Admins control available channels in the Power Platform admin center.18

Microsoft Copilot Studio architecture: components by layer and how they connectCopilot & businessappsAgents & automationAI models & platformData & analyticsDeveloper toolsSecurity, identity &governanceChannels (Teams and Microsoft Copilot, SharePoint, web, WhatsApp, Facebook, Azure Bot Service): Where employees and customers talk to the agentChannels (Teams andMicrosoft Copilot,…Agents: Follow instructions, use knowledge and tools, and reason about next stepsAgentsWorkflows and agent flows: Structured automations that can call agents and tools with human reviewWorkflows and agent flowsHarnesses (GitHub Copilot, Customer Conversation, standard): Engine that sets reasoning style, capabilities and billingHarnesses (GitHub Copilot,Customer Conversation,…Knowledge sources: Company content and data the agent grounds its answers inKnowledge sourcesCopilot Studio web app: Low-code studio at copilotstudio.microsoft.com for building and managing agents and workflowsCopilot Studio web appConnectors, MCP servers and tools: Let agents read from and act in business systemsConnectors, MCP servers andtoolsPower Platform admin center and Entra ID: Environments, channel controls, billing policies and authenticationPower Platform admin centerand Entra ID
Diagram as a list
  1. Copilot & business apps

    • Channels (Teams and Microsoft Copilot, SharePoint, web, WhatsApp, Facebook, Azure Bot Service)Where employees and customers talk to the agent
  2. Agents & automation

    • AgentsFollow instructions, use knowledge and tools, and reason about next steps
    • Workflows and agent flowsStructured automations that can call agents and tools with human review
  3. AI models & platform

    • Harnesses (GitHub Copilot, Customer Conversation, standard)Engine that sets reasoning style, capabilities and billing
  4. Data & analytics

    • Knowledge sourcesCompany content and data the agent grounds its answers in
  5. Developer tools

    • Copilot Studio web appLow-code studio at copilotstudio.microsoft.com for building and managing agents and workflows
    • Connectors, MCP servers and toolsLet agents read from and act in business systems
  6. Security, identity & governance

    • Power Platform admin center and Entra IDEnvironments, channel controls, billing policies and authentication
Components and connections as documented by Microsoft.8

Capabilities

  • Agents built in natural language1

    Makers describe what the agent should do, connect knowledge and tools, and test before publishing; Copilot Studio can also propose a mix of agents and workflows from a business goal.

    Why it matters: Lowers the skill bar for building useful agents.

    Limits: Building with natural language across agents and workflows is marked preview.

  • Harness choice1

    GitHub Copilot harness for reasoning-heavy work, Customer Conversation harness for customer voice and chat with sub-agents and human handoff, standard harness for topic-based conversations.

    Why it matters: Teams pick predictability or flexibility per agent.

    Limits: The harness changes billing, features and capabilities, so licensing differs by harness.

  • Workflows and agent flows1

    Drag-and-drop workflows combine agents, tools and logic with testing and human-in-the-loop steps; agent flows can run alone or as tools that return results to an agent.

    Why it matters: Mixes predictable automation with AI reasoning in one solution.

    Limits: Complex integrations still depend on available connectors or custom connectors.

  • Connectors, MCP servers and model choice24

    Microsoft lists more than 1,500 prebuilt connectors and MCP servers, plus model selection that includes models from Foundry.

    Why it matters: Agents can act in business systems without custom code.

    Limits: Premium connectors need the standalone subscription; the product page still refers to Azure AI Foundry, the former name of Microsoft Foundry.

  • Multi-channel publishing8

    Publish to Teams and Microsoft Copilot, SharePoint, WhatsApp, websites, mobile apps, Facebook and Azure Bot Service channels.

    Why it matters: One agent can serve employees and customers in different places.

    Limits: Users cannot send attachments to the agent in any channel unless an Azure Bot Service skill handles them, and Teams shows at most six suggested actions.

  • Governance and cost controls2

    Analytics, evaluations, agent inventory, role-based access, environment controls, audit logs to Purview and budgets for AI credit use.

    Why it matters: IT can limit agent sprawl and spending.

    Limits: Capacity is enforced monthly, unused credits do not carry over, and exceeding purchased capacity can lead to service denial.

Practical use cases

New hires flood HR with the same questions about policies and forms.
Approach
Build an onboarding agent in Copilot Studio grounded in the HR SharePoint site and publish it to Teams and Microsoft Copilot.
Role of Microsoft Copilot Studio
Copilot Studio hosts the agent and its knowledge; Microsoft cites this scenario as a fit.
Data, infrastructure and skills
Curated HR content and a Copilot Studio licensing path.
Type of benefit
Faster answers for employees
Caveats
Answers depend on content quality; for Copilot-licensed users some uses are zero-rated, for others credits apply.
First step
Collect the 20 most common HR questions and test the agent against them.
A contact center wants a voice and chat agent that resolves routine requests and hands harder ones to staff.
Approach
Use the Customer Conversation harness with defined topics and an escalation path to human representatives.
Role of Microsoft Copilot Studio
Copilot Studio runs the primary agent and sub-agents and performs the handoff.
Data, infrastructure and skills
Channel setup for voice or chat and an escalation destination.
Type of benefit
Reduced handling load
Caveats
Billing and features depend on the harness; test containment before broad launch.
First step
Pick one request type, such as order status, and build a pilot agent.
An approval process spans email, a spreadsheet and a line-of-business system.
Approach
Model it as a Copilot Studio workflow that uses connectors for each system, an agent step for judgment calls and a human approval step.
Role of Microsoft Copilot Studio
Copilot Studio combines automation and AI reasoning with human review.
Data, infrastructure and skills
Connectors for the systems involved.
Type of benefit
Process automation
Caveats
Systems without connectors need custom connectors.
First step
Map the current steps and check connector availability for each system.

Who uses it

Works with

Optional integration

  • Microsoft 365 CopilotAgents publish to Microsoft Copilot; a Copilot license covers GitHub Copilot harness agents and zero-rates some use in Copilot Chat, Teams and SharePoint.
  • Microsoft TeamsTeams and Microsoft Copilot is a primary publishing channel.
  • Microsoft Power PlatformUses Power Platform environments, admin center, billing policies and connectors.
  • Microsoft FoundryThe product page lists model selection that includes Foundry models.

Complementary tools

  • Microsoft 365 CopilotA Copilot license lets users build Copilot Studio agents on the GitHub Copilot harness, and licensed use of agents in Copilot Chat, Teams or SharePoint is zero-rated for some features.
  • Microsoft Agent 365Agent 365 inventories Copilot Studio agents and extends Entra network controls to them.
  • Microsoft Copilot ChatAgents built in Copilot Studio can be used in Copilot Chat; agents with work data run pay-as-you-go.

Same family

Alternative approaches

Optional integration for

  • Microsoft Dragon CopilotMicrosoft named Copilot Studio's healthcare agent service as a foundation for partners building solutions that surface in Dragon Copilot (October 2025).
  • Microsoft Dynamics 365Custom agents for Dynamics 365 are built with Copilot Studio.
  • Microsoft PurviewCopilot Studio is one of the Copilot experiences Purview covers.
  • Microsoft TeamsTeams and Microsoft Copilot is a publishing channel for Copilot Studio agents.
  • Windows 365Copilot Studio computer use can run on a Cloud PC pool from Windows 365 for Agents (preview).
  • Azure AI SearchA Foundry IQ knowledge base can be connected to a Copilot Studio agent, and Copilot Studio integrates with Azure AI Search.
  • Azure DevOpsCopilot Studio is a supported environment for the remote Azure DevOps MCP Server.

Relationship labels follow Microsoft's documentation. "Alternative approaches" does not mean one is better: each profile says when it fits.

Compared with

Getting started

  1. Choose a licensing path

    Decide between a prepaid Copilot Credit pack, pay-as-you-go through Azure, a prepurchase plan, or using Microsoft 365 Copilot licenses for GitHub Copilot harness agents.

    Check: Makers can sign in at copilotstudio.microsoft.com and the environment shows available capacity.

  2. Set up governance

    In the Power Platform admin center, assign the Copilot Studio authors security group, link a billing policy if using pay-as-you-go, and review allowed channels.

    Check: Only intended makers can create agents and blocked channels show as disabled.

  3. Build and test a first agent

    Describe the agent, add one knowledge source and one tool, and test in the test panel.

    Check: The agent answers test questions with citations to the connected knowledge.

  4. Publish to yourself, then widen

    Publish, open the agent in Teams for your own use, then make it available to others.

    Check: The agent appears in Teams and Microsoft Copilot for the pilot group.

  5. Estimate consumption

    Use the Copilot Studio agent usage estimator to forecast Copilot Credits.

    Check: You have a monthly credit estimate before a wide rollout.

Official resources

Deciding whether you need it?

Three free tools apply Microsoft's own published guidance to your situation, and say when Microsoft Copilot Studio or any new technology is not needed.

Sources

Each statement above links to the source it comes from. Labels say who reported it.

  1. Overview: Microsoft Copilot Studio (opens in a new tab) Microsoft · Vendor-reported · link checked 27 Oct 2026
  2. Standard harness licensing: Microsoft Copilot Studio (Microsoft Learn) (opens in a new tab) Microsoft · Vendor-reported
  3. Custom engine agents for Microsoft 365 (opens in a new tab) Microsoft · Vendor-reported
  4. Microsoft Copilot Studio (product page) (opens in a new tab) Microsoft · Vendor-reported · link checked 27 Oct 2026
  5. Agent Builder overview (Microsoft Learn) (opens in a new tab) Microsoft · Vendor-reported
  6. Technology plan for AI agents (Cloud Adoption Framework) (opens in a new tab) Microsoft · Vendor-reported · link checked 27 Oct 2026
  7. Microsoft Agent Framework overview (opens in a new tab) Microsoft · Vendor-reported · link checked 27 Oct 2026
  8. Publish and deploy your agent: Microsoft Copilot Studio (Microsoft Learn) (opens in a new tab) Microsoft · Vendor-reported

Fill out the form below to request your copy.

Name(Required)