--- title: "What Are Microsoft Copilot Studio Agents?" url: "https://atlan.com/know/ai-agent/ai-agent-applications/what-are-microsoft-copilot-studio-agents/" description: "See how Microsoft Copilot Studio agents get built, published, priced in Copilot Credits, and governed through Microsoft Agent 365 and Purview in 2026." author: "Emily Winks" author_role: "Data Governance Expert" published: "2026-08-12" updated: "2026-08-12" --- --- Microsoft Copilot Studio is Microsoft's low-code platform for building and publishing custom AI agents into Teams, SharePoint, Microsoft 365, and public websites. It prices usage in Copilot Credits, from $0.01 each to $200 for a 25,000-credit monthly pack, not per seat. Atlan's Enterprise Data Graph and **MCP Server** give the definitions those agents rely on somewhere to stay certified once more than one team is building on the platform. A Copilot Studio agent is only as consistent as the knowledge source feeding it, and most enterprises end up with more than one, built independently across departments. - Agent Builder: quick, personal or small-team agents from natural language - Copilot Studio: the fuller platform for org-wide, multi-step agents - Microsoft Agent 365: governance and identity, once you have more than a few agents - The Store Operations Agent: one prebuilt, preview-stage retail template, not the general product | What Microsoft Copilot Studio Is | | | ----- | ----- | | **What it is** | Microsoft's low-code platform for building and publishing custom AI agents | | **Key benefit** | One-click upgrade from Agent Builder to a governed, org-wide agent | | **Best for** | Microsoft 365 and Teams-native enterprises automating support, HR, and line-of-business work | | **Governance layer** | Microsoft Agent 365 (GA May 1, 2026) plus Microsoft Purview data loss prevention | | **Pricing model** | Copilot Credits: $200/month per 25,000 credits, or $0.01/credit pay-as-you-go | | **Core components** | Agent Builder, Copilot Studio, knowledge sources, 1,400+ connectors, MCP support, Agent 365 | --- ## What is Microsoft Copilot Studio? Microsoft Copilot Studio is a low-code platform, built on Microsoft's Power Platform, for creating custom [AI agents](https://atlan.com/know/ai-agent/what-is-an-ai-agent/) that take multi-step action inside business workflows, the distinction that separates a genuine agent from [an enterprise copilot](https://atlan.com/know/ai-agent/what-is-an-enterprise-copilot/) that only answers on request. It sits alongside, not inside, Microsoft's other AI products: Microsoft 365 Copilot handles everyday productivity, Security Copilot serves security teams, Dynamics 365 Copilot is scoped to CRM and ERP. Copilot Studio is the maker platform underneath all of them, where a developer builds a custom agent rather than using one Microsoft shipped by default. [Microsoft's 2026 release-wave documentation](https://learn.microsoft.com/en-us/power-platform/release-plan/2026wave1/microsoft-copilot-studio/) describes Copilot Studio as the fully managed platform organizations use to build, govern, and scale agents, putting governance in the product's stated purpose, not as an afterthought. The practical question most teams have isn't whether Copilot Studio can build an agent. It's what happens once ten teams have each built one, the same scaling question covered in [how enterprises use AI agents](https://atlan.com/know/ai-agent/how-enterprises-use-ai-agents/) generally. | Version | Where it fits | | ----- | ----- | | Agent Builder (Microsoft 365 Copilot) | Quick, personal, or small-team agents from natural language, no separate license | | Copilot Studio | Org-wide agents with connectors, knowledge sources, and multi-step actions | | Microsoft Agent 365 | Identity, registry, and governance layer that tracks agents built anywhere, including Copilot Studio | | Microsoft 365 Copilot, Security Copilot, Dynamics 365 Copilot | Separate, purpose-built Microsoft products, not built on the same maker surface | --- ## How do you build and publish an agent in Copilot Studio? Building an agent in Copilot Studio means connecting it to knowledge sources and connectors first, then choosing where it gets published. [Microsoft's own knowledge-sources documentation](https://learn.microsoft.com/en-us/microsoft-copilot-studio/agents-experience/knowledge-sources-overview) lists SharePoint, OneDrive, Microsoft Teams chats and channels, and Dataverse as native sources, alongside structured data from platforms like Snowflake and Databricks. None of those sources are exclusive to one agent: the same SharePoint site can ground three different agents built by three different teams, each maintaining its own separate copy of "how we describe this," the exact scenario [how to build a knowledge base for AI agents](https://atlan.com/know/ai-agent/data-for-ai/how-to-build-knowledge-base-for-ai-agents/) warns against treating as a one-time setup step. Copilot Studio ships with [more than 1,400 external connectors](https://learn.microsoft.com/en-us/microsoft-copilot-studio/advanced-connectors), plus Work IQ for workflow-tuned agents. [Model Context Protocol support reached general availability](https://www.microsoft.com/en-us/microsoft-copilot/blog/copilot-studio/multi-agent-orchestration-maker-controls-and-more-microsoft-copilot-studio-announcements-at-microsoft-build-2025/) in 2025, meaning an agent can call any [MCP](https://atlan.com/know/mcp-vs-function-calling/) server, not only Microsoft's own, the same open protocol Atlan's own [MCP server](https://atlan.com/know/ai-agent/how-to-build-mcp-servers-for-enterprise-data/) speaks. One prebuilt example makes this concrete. Microsoft ships a [Store Operations Agent template, currently in preview](https://learn.microsoft.com/en-us/industry/retail/store-operations-template-copilot-studio/overview-store-operations-template-copilot-studio/), scoped narrowly to retail frontline work: store policies, inventory, and shipping status, wired to a retailer's systems through preexisting or custom integrations. It's one narrow starting point Microsoft happens to have published, not a description of what a Copilot Studio agent generally does. Publishing is the last step: agents deploy directly into Teams, SharePoint, and Microsoft 365 Copilot, or into a public website. Each target is a different surface with its own users, which means the same agent can answer identically only if its knowledge sources stay [fresh](https://atlan.com/know/ai-agent/context-freshness/) everywhere it's published. --- ## How is Copilot Studio different from Agent Builder in Microsoft 365 Copilot? The distinction between Agent Builder and Copilot Studio is the single most-repeated point of confusion in Microsoft's own practitioner community, and it comes down to audience and depth, not capability tiers. [Microsoft's own comparison documentation](https://learn.microsoft.com/en-us/microsoft-365/copilot/extensibility/copilot-studio-experience) frames Agent Builder as the tool for quickly creating an agent for yourself or a small team, while Copilot Studio is built for enterprise-scale automation: internal systems, triggered actions, APIs, and business rules. | Aspect | Agent Builder (Microsoft 365 Copilot) | Copilot Studio | | ----- | ----- | ----- | | Typical builder | Any licensed Microsoft 365 Copilot user | Maker, admin, or developer | | Typical audience | Themselves or a small team | A department, the whole org, or external customers | | Setup | Natural language, no separate license | Low-code, connectors, knowledge sources, publishing controls | | Multi-step actions | Limited | Full support: connectors, APIs, business rules | | Upgrade path | One-click promotion into Copilot Studio | N/A (the destination) | The upgrade path matters more than the table implies. An agent isn't rebuilt from scratch when it outgrows Agent Builder; it promotes in one click, which is why agent counts in a tenant tend to grow faster than IT expects: one person's experiment is now a governed, org-facing agent with its own knowledge source to maintain. Build your AI context stack Get the blueprint for implementing context graphs across your enterprise, from metadata foundation to agent orchestration, with practical steps for 2026. Get the Stack Guide --- ## What does a Copilot Studio agent cost? Copilot Studio prices usage by the credit, pooled at the tenant level, not by the seat, so cost tracks what agents do, not how many employees can talk to them. [According to CloudZero's 2026 pricing analysis](https://www.cloudzero.com/blog/copilot-studio-pricing/), Microsoft renamed the usage currency from "messages" to "Copilot Credits" on September 1, 2025, without changing the pack size or meter rate. Two consumption paths exist side by side, confirmed on [Microsoft's own pricing page](https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/copilot-studio) and [Azure's pay-as-you-go page](https://azure.microsoft.com/en-us/pricing/details/copilot-studio/): a prepaid pack at $200/month for 25,000 credits, or pay-as-you-go at $0.01/credit. M365 Copilot-licensed employees' internal agents draw no additional credits. What drives the bill is what an interaction does: a classic answer burns roughly one credit, generative roughly two, tenant-graph grounding roughly ten, and a fully autonomous action 25 or more. | Consumption path | Unit cost | Best for | | ----- | ----- | ----- | | Internal agents (M365 Copilot-licensed users) | No additional credit cost | Employee-only agents with existing licenses | | Prepaid capacity pack | $200/month per 25,000 credits | Predictable, budgeted usage | | Pay-as-you-go (Azure) | $0.01/credit | Variable or bursty usage | The credit cost of an agent scales with its autonomy, not its popularity, the same math behind [the cost to run AI agents at scale](https://atlan.com/know/ai-agent/cost-to-run-ai-agents-at-scale/) on any platform. An agent that only answers questions is cheap. One taking autonomous actions costs 25 times more per interaction, exactly the kind that needs governance once it's [running in production](https://atlan.com/know/ai-agent/ai-agent-scaling-in-production/), not a pilot. --- ## What governance and security controls does Microsoft provide for Copilot Studio agents? Microsoft has built a real governance stack around Copilot Studio, the territory covered in [AI agent risks and guardrails](https://atlan.com/know/ai-agent-risks-guardrails/) generally. [Microsoft Purview](https://learn.microsoft.com/en-us/purview/ai-copilot-studio) extends data loss prevention to Copilot Studio agents: data policies, sensitive-item detection, and admin controls in the Power Platform admin center. [Microsoft Agent 365 reached general availability on May 1, 2026](https://holgerimbery.blog/agent-365-ga), a $15-per-user-per-month add-on folded into Microsoft 365 E7. It's a control plane for what shows up once an org has more than a handful of agents: every agent gets a Microsoft Entra Agent ID, and IT gets a central [agent registry](https://atlan.com/know/ai-agent/ai-agent-identity/) covering agents built anywhere, Copilot Studio included. Microsoft built that control plane for a well-documented pattern. [Gartner projects more than 40% of agentic AI projects will be canceled by the end of 2027](https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027), citing escalating costs and inadequate risk controls. Anushree Verma, Senior Director Analyst at Gartner: "Most agentic AI projects right now are early stage experiments or proof of concepts that are mostly driven by hype and are often misapplied." A 2026 Gartner forecast goes further, [projecting 40% of enterprises will demote or decommission autonomous agents by 2027](https://www.forbes.com/sites/janakirammsv/2026/06/09/microsoft-makes-governance-the-gate-for-enterprise-ai-agents/) after governance failures surface in production, the same [accuracy](https://atlan.com/know/ai-agent/ai-agent-accuracy/) gap that traces to context, not the model. The registry answers how many agents exist and who owns each one. It doesn't answer whether they agree with each other, the [observability](https://atlan.com/know/ai-agent-observability/) question a registry alone can't settle. Find your context gap Most agent rollouts stall on data readiness and consistency, not the reasoning model. See where your knowledge sources have gaps before you scale past a pilot. Calculate Your Gap --- ## What should a data leader consider before scaling Copilot Studio agents? Scaling past one Copilot Studio agent surfaces a consistency problem a single agent never exposes, since it only has to agree with itself. First: knowledge-source freshness. A SharePoint site or Dataverse table is maintained on whatever schedule its owner keeps, and an agent grounded in a stale copy answers confidently and wrong, the failure mode [context quality testing](https://atlan.com/know/ai-agent/context-quality-testing-for-ai-agents/) exists to catch. Second: cross-tool consistency, where the math changes as agent count grows. One team's agent might define "active customer" one way; another's, built on a different knowledge source, might define it differently, and nothing in Copilot Studio forces the two to reconcile. [How AI agents work](https://atlan.com/know/ai-agent/ai-agent-architecture-explained/) explains this isn't a reasoning-layer bug: two agents given different definitions of the same term act on them consistently, just not consistently with each other. [Metadata management for AI](https://atlan.com/know/ai-agent/data-for-ai/metadata-management-for-ai/) is the discipline that keeps those definitions from drifting apart. From one agent to many: where Copilot Studio's governance stops. Source: Atlan Third, what [Agent 365's own design already concedes](https://atlan.com/know/ai-agent-governance/): a registry tracking identity and access is necessary but not sufficient. Knowing which agents exist differs from knowing whether they agree, and Microsoft's stack solves the first, not the second, the same wall enterprises hit [making AI agents context-aware](https://atlan.com/know/ai-agent/how-to-make-ai-agents-context-aware/) on [enterprise-ready AI agents](https://atlan.com/know/ai-agent/enterprise-ready-ai-agents/), regardless of platform. It's [why AI agents need an enterprise context layer](https://atlan.com/know/ai-agent/who-owns-ai-governance-caio-vs-cdo-vs-cto/) that spans platforms. --- ## How Atlan extends context beyond a single Copilot Studio agent **The gap.** Copilot Studio makes it fast to connect a knowledge source and publish an agent. It doesn't give every agent a shared, certified definition of a business term, because each agent's source is scoped to whatever its builder connected. **Atlan's approach.** Atlan's own [MCP server](https://atlan.com/know/ai-agent/how-to-build-mcp-servers-for-enterprise-data/) speaks the same open protocol Copilot Studio supports natively. The [Enterprise Data Graph](https://atlan.com/know/ai-agent/knowledge-graph-for-ai-agents/) carries lineage and certification across whatever feeds a knowledge source, Microsoft-native or not, and a [semantic layer](https://atlan.com/know/ai-agent/semantic-layer-for-ai-agents/) resolves a term like "revenue" once, the [context layer](https://atlan.com/know/what-is-context-layer/) idea applied to one platform's agents. **The outcome.** [Data quality for AI agents](https://atlan.com/know/data-for-ai/data-quality-for-ai-agent/) research finds agents on well-governed, consistently-defined data behave predictably, because the context feeding every agent agreed with itself first. | Aspect | Copilot Studio alone | + governed context layer (e.g. Atlan) | | ----- | ----- | ----- | | Knowledge sources | Scoped per agent (SharePoint, Dataverse, connectors) | Same sources, plus one certified definition layer every agent reads from | | Governance | Agent 365 registry, Entra Agent ID, Purview DLP | Adds lineage, ownership, and certification across every connected system | | Cross-platform consistency | Not addressed; each agent stands alone | One semantic layer reused by agents on any platform, Copilot Studio included | | Protocol reach | MCP-native, 1,400+ connectors | Extended by an external, governed MCP server | See context agents live Watch how a governed context layer keeps agents on different platforms answering from the same certified definitions, in real time. Watch a Live Demo --- ## What does Copilot Studio's own governance push reveal about enterprise AI context? Microsoft didn't build Agent 365 and extend Purview into Copilot Studio because governance is fashionable. It built them because a platform that makes agent creation this easy will produce agent sprawl faster than any manual review process can track, and Microsoft's own release notes treat that sprawl as the expected outcome, not an edge case. The registry, the identity records, the data loss prevention policies: all of it is Microsoft solving for "we don't know how many agents exist or what they can touch," which is the right first problem to solve. It isn't the last one. An enterprise that knows exactly how many Copilot Studio agents it has, and who owns each one, can still end up with agents that quietly disagree about what a customer, a product, or a dollar of revenue means, because Agent 365's registry was never designed to check for that. Atlan extends the same [context engineering](https://atlan.com/know/ai-agent/context-engineering/context-engineering-techniques-ai-agents/) discipline across whatever an enterprise runs, Copilot Studio agents, agents built elsewhere, or both, so the question worth asking isn't whether Copilot Studio's governance stack is thorough. It's whether every agent reading from it is still telling the same story. Book a Demo --- ## FAQs about Microsoft Copilot Studio agents ### 1. What is Microsoft Copilot Studio? Microsoft Copilot Studio is a low-code platform for building and publishing custom AI agents into Teams, SharePoint, Microsoft 365, and public websites, distinct from Microsoft 365 Copilot, Security Copilot, and Dynamics 365 Copilot. ### 2. What is the difference between Copilot Studio and Agent Builder in Microsoft 365 Copilot? Agent Builder quickly creates a personal or small-team agent from natural language, no separate license. Copilot Studio is for enterprise-scale agents connecting to internal systems and business rules, and an Agent Builder agent upgrades into Copilot Studio in one click. ### 3. How much does Copilot Studio cost? Copilot Studio prices usage in Copilot Credits: prepaid packs run $200 per month for 25,000 credits, or $0.01 per credit pay-as-you-go via Azure. M365 Copilot-licensed employees' internal agents draw no additional credits. ### 4. Can Copilot Studio agents use MCP (Model Context Protocol)? Yes. MCP support reached general availability in Copilot Studio in 2025, so an agent can call any MCP server as a context source, not only Microsoft's own connectors. ### 5. What is Microsoft Agent 365, and how does it relate to Copilot Studio? Microsoft Agent 365 is a separate governance and identity control plane, GA since May 1, 2026, priced at $15 per user per month. It gives every agent, Copilot Studio included, an Entra Agent ID and a registry tracking who owns each one. ### 6. Is the Store Operations Agent the same thing as Copilot Studio? No. It's one prebuilt, preview-stage template built on Copilot Studio, scoped to retail tasks like inventory and shipping lookups. Copilot Studio is the general-purpose platform; the template is one example built with it. ### 7. Where can a Copilot Studio agent be published? Into Microsoft Teams, SharePoint, Microsoft 365 Copilot, or a public website. The same agent can publish to more than one target, so its knowledge sources need to stay current everywhere it's deployed. ### 8. How does Atlan connect to agents built in Copilot Studio? Atlan's own MCP server speaks the same Model Context Protocol Copilot Studio supports natively, so a Copilot Studio agent can call it as an external, governed context source for definitions that live outside its native knowledge sources. --- ## Sources 1. [Overview of Microsoft Copilot Studio, Microsoft Learn (2026)](https://learn.microsoft.com/en-us/power-platform/release-plan/2026wave1/microsoft-copilot-studio/) 2. [Available knowledge sources for agents, Microsoft Learn (2026)](https://learn.microsoft.com/en-us/microsoft-copilot-studio/agents-experience/knowledge-sources-overview) 3. [Use connectors in Copilot Studio agents, Microsoft Learn (2026)](https://learn.microsoft.com/en-us/microsoft-copilot-studio/advanced-connectors) 4. [Copilot Studio announcements at Microsoft Build 2025, Microsoft (2025)](https://www.microsoft.com/en-us/microsoft-copilot/blog/copilot-studio/multi-agent-orchestration-maker-controls-and-more-microsoft-copilot-studio-announcements-at-microsoft-build-2025/) 5. [Store Operations Agent in Copilot Studio (preview), Microsoft Learn (2026)](https://learn.microsoft.com/en-us/industry/retail/store-operations-template-copilot-studio/overview-store-operations-template-copilot-studio/) 6. [Choose between Agent Builder and Copilot Studio, Microsoft Learn (2026)](https://learn.microsoft.com/en-us/microsoft-365/copilot/extensibility/copilot-studio-experience) 7. [Microsoft Copilot Studio Pricing In 2026, CloudZero (2026)](https://www.cloudzero.com/blog/copilot-studio-pricing/) 8. [Microsoft Purview for Copilot Studio, Microsoft Learn (2026)](https://learn.microsoft.com/en-us/purview/ai-copilot-studio) 9. [Microsoft Agent 365 Is Now Generally Available, Holger Imbery (2026)](https://holgerimbery.blog/agent-365-ga) 10. [Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027, Gartner (2025)](https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027) 11. [Microsoft Makes Governance The Gate For Enterprise AI Agents, Forbes (2026)](https://www.forbes.com/sites/janakirammsv/2026/06/09/microsoft-makes-governance-the-gate-for-enterprise-ai-agents/) 12. [Microsoft 365 Copilot Pricing: Copilot Studio, Microsoft (2026)](https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/copilot-studio) 13. [Copilot Studio Pay As You Go Pricing, Microsoft Azure (2026)](https://azure.microsoft.com/en-us/pricing/details/copilot-studio/)