What is Microsoft Copilot Studio? An honest guide for businesses.

Microsoft uses the word Copilot for a whole family of products, which makes it easy to buy the wrong one. Copilot Studio is the one you build with. Microsoft 365 Copilot is the assistant your employees use. Mixing them up leads to licences nobody needs and agents that never leave the test panel.

This guide explains what Copilot Studio is, what you can build with it, what it costs and where it falls short. Every feature and price below was checked against the vendors’ own documentation and pricing pages in October 2026. Microsoft renames products and changes prices often, so treat the figures as a snapshot and confirm them before you sign anything. At the end you will find an honest comparison with n8n, the platform we build our client automations on.

What is Microsoft Copilot Studio, in plain terms?

Microsoft Copilot Studio is a graphical, low-code studio for building and managing AI agents and workflows. That is Microsoft’s own definition on Microsoft Learn. You build agents and workflows, connect them to your organisation’s data and systems, and publish them to the channels where people already work.

Copilot Studio is the successor to Power Virtual Agents, which Microsoft folded into it. Since then it has grown well beyond chatbots. In September 2026 Microsoft added apps in public preview, so you can now build an interface, a workflow and an agent in one place.

You work in it through a web app at copilotstudio.microsoft.com. You can start from a blank canvas, but the quickest start is a plain-language description of the goal. Copilot Studio drafts the agent or workflow, and you refine it from there.

How Copilot Studio differs from Microsoft 365 Copilot

The short version: Microsoft 365 Copilot helps people work, Copilot Studio is where you build what does the work.

Agents you create in Agent Builder can be exported to Copilot Studio when they need more control, governance or channels.

What can you build with Copilot Studio?

Copilot Studio has three main building blocks: agents, workflows and agent flows, with apps added in preview. You can use each one on its own or combine them. A workflow can call an agent to handle one step, and an agent can call a flow as a tool.

Agents

An agent is an AI assistant that follows your instructions, draws on the knowledge you connect and uses tools to take action. Agents work in multiple languages across Teams, Microsoft 365 Copilot, websites, mobile apps and messaging channels such as Facebook Messenger. Some agents can get their own account, so they can work proactively on shared processes such as onboarding a new employee.

Microsoft’s own examples include an IT support agent that handles ticket intake and triage, a recruitment assistant that screens and ranks candidates, a contract review agent that flags risks and a balance sheet reconciliation agent that detects variances.

Topics

Topics are the scripted part of a conversation. On the standard harness, a topic is a piece of dialogue you design with trigger phrases, questions, conditions and steps. The agent uses natural language understanding to match a request to the best topic. When no topic fits, it can still generate an answer from its connected knowledge.

Use topics where the answer must be exactly right every time: a returns procedure, a booking flow, an escalation to a human.

Tools

Tools let an agent act instead of only talk. Older guides call them actions. You can add:

• Prebuilt connectors to Microsoft and non-Microsoft services, for example to send an email through Outlook or post a message in Teams.
• Custom connectors to your own systems.
• Agent flows, predefined sequences of steps the agent can trigger.
• Prompts, single model calls for a specific task.
• REST APIs, by selecting the endpoints the agent may call.
• MCP servers, through the Model Context Protocol.
• Computer use, which lets the agent operate websites and desktop apps through their interface, clicking buttons and filling in fields.

With generative orchestration switched on (the default), the agent picks the right tool itself, based on each tool’s name and description. Microsoft supports up to 128 tools per agent, but recommends 25 to 30 for the best results.

Knowledge

Knowledge grounds an agent’s answers in your own content. Copilot Studio supports public websites (searched through Bing), uploaded documents, SharePoint, Dataverse and enterprise data indexed through Microsoft Copilot connectors. For SharePoint and Dataverse, the agent signs in with the user’s own Microsoft Entra ID, so people only get answers from content they are already allowed to see.

Connectors and the Power Platform

Copilot Studio is part of Microsoft’s Power Platform, next to Power Automate and Power Apps, and they share one connector library. Microsoft lists more than 1,500 prebuilt connectors, including Salesforce, Dropbox and Google services, and you can build custom connectors for anything else.

That shared foundation matters. Agents can store structured data in Dataverse, reach on-premises data through Power Platform connectors and are managed from the Power Platform admin center. There, IT sets the data policies that decide which connectors and channels makers may use. If your company already runs Power Automate flows and Power Apps, Copilot Studio fits into an environment, permission model and governance setup your IT team already knows.

Workflows and agent flows

Workflows are automations you build in a drag-and-drop designer, or by describing them so Copilot Studio can generate and validate them. Each step can combine agents, tools and fixed logic, with built-in testing and human approval steps.

Agent flows are the older, established format. They feel like Power Automate, run as standalone automations or as a tool for an agent, and can include prompts and human review steps.

Harnesses: the engine underneath

Everything you build runs on a harness, the runtime that decides when to call the model, which tools to use and how the work is billed. There are three:

Despite the name, the GitHub Copilot harness is not the GitHub Copilot service. Microsoft states that customer data is not sent to or processed by GitHub Copilot when agents run in Copilot Studio.

Model choice

OpenAI models are the default for new agents. Since September 2025, admins can also enable Anthropic’s Claude models for orchestration and prompts. For custom scenarios, Microsoft points to the more than 11,000 models in Azure AI Foundry, which are billed separately from Copilot Credits.

How much does Copilot Studio cost in October 2026?

Copilot Studio is billed in Copilot Credits, a usage currency that replaced “messages” on 1 September 2025. Internal agents used by people with a Microsoft 365 Copilot licence are largely included. Everything else is paid per credit, prepaid or pay-as-you-go. These are Microsoft’s US list prices as of October 2026, from the Copilot Studio pricing page, with the euro amounts from Microsoft’s Dutch page where it shows them.

Smaller organisations, up to 300 users, can choose Microsoft 365 Copilot Business instead. It is an add-on to an eligible Microsoft 365 business plan, listed from $21 per user per month, with a first-year promotional price of $18 on annual plans bought between 1 July and 31 December 2026.

Two details in the fine print deserve attention:

• Unused credits do not roll over. Microsoft enforces prepaid capacity per month.
• Overspending switches agents off. When a tenant reaches 125% of its prepaid capacity, custom agents are disabled and users see a message such as “This agent is currently unavailable. It has reached its usage limit.” Linking a pay-as-you-go meter prevents this, at pay-as-you-go prices.

What one credit buys

On the standard harness, Microsoft charges a fixed number of credits per activity. From the Copilot Credits billing rates:

A reasoning model adds a premium of 10 credits per 1,000 tokens on top of the activity rate.

For employees with a Microsoft 365 Copilot licence, these activities cost nothing when the agent runs under their identity inside Microsoft 365. There are exceptions: agent flows are only free when an agent triggers them for a licensed user, and computer-using agents are not covered by the licence.

The GitHub Copilot harness works differently. It charges credits for model tokens, tools (including knowledge and MCP) and the harness itself, and billing starts while you build. Generating an agent from a description, testing it and running evaluations all consume credits.

A worked example: a support agent on your website

Microsoft’s own billing example is a customer support agent on a website that answers questions from return policies and product manuals. An average conversation uses four classic answers and two generative answers, and 900 customers use it every day.

• Per day: (4 x 1 + 2 x 2) x 900 = 7,200 credits.
• Per 30-day month: 216,000 credits.
• On pay-as-you-go at $0.01 per credit: about $2,160 a month.
• With prepaid packs: nine packs, $1,800 a month.

For a busy service channel, that can be reasonable. The lesson is that the bill grows with every answer and every action, so model your volumes before you launch. Microsoft offers a Copilot Studio agent usage estimator for exactly this.

Costs outside the credit price

• Power Automate cloud flows use Power Automate licensing, not Copilot Credits.
• Dataverse storage above the included capacity is billed separately. On pay-as-you-go, database storage above 1 GB per environment costs $48 per GB per month, according to Microsoft’s pay-as-you-go meters.
• Your own models from Azure AI Foundry are billed separately.
• Administration time for environments, data policies and capacity. Somebody has to own it.

What are the strengths of Copilot Studio?

Copilot Studio’s biggest strength is that it lives where Microsoft 365 users already work, under the security model IT already runs.

• No new app to roll out. Agents appear in Teams and Microsoft 365 Copilot, so adoption starts in a familiar place.
• Security and governance from the Microsoft stack. Microsoft Entra ID, Defender, conditional access, audit logs in Microsoft Purview and per-environment data policies come built in. Microsoft states that Copilot Studio supports data residency alignment and major compliance standards.
• Included usage for licensed staff. If your employees already have Microsoft 365 Copilot, many internal agent interactions cost nothing extra.
• Low-code with real depth. Natural-language building for speed, topics for control, evaluations and a Review panel before publishing, and release pipelines with approval gates for development, test and production.
• A mature connector library shared with Power Automate and Power Apps.
• Model choice between OpenAI and Anthropic models.

What are the limits of Copilot Studio?

The main limits are licensing complexity, costs that scale with usage and a firm tie to Microsoft’s cloud.

• Licensing is hard to read. Three harnesses with different billing, credit packs, pay-as-you-go, pre-purchase units and inclusion rules that depend on the licence, the trigger and the channel. Microsoft itself refers readers to a separate licensing guide for the complete picture.
• Costs scale with every interaction. Answers, actions and reasoning tokens are all metered. That is fine for internal use under Microsoft 365 Copilot, but customer-facing agents and high-volume automations need a careful budget.
• You pay while you build. On the GitHub Copilot harness, building and testing consume credits too.
• Microsoft cloud only. You cannot run Copilot Studio on your own servers. Your agents, flows and Dataverse tables live on Microsoft’s platform, and moving away later means rebuilding them.
• Strongest inside Microsoft. Connectors reach other vendors, but features such as tenant graph grounding and user-level permissions on knowledge assume your data lives in Microsoft 365.
• It changes fast. Power Virtual Agents became Copilot Studio, actions became tools, messages became Copilot Credits, harnesses arrived and apps are in preview. Good for features, harder for documentation, training and budgets.
• Small channel gaps. On the standard harness, users cannot upload attachments in the chat, on any channel.

Copilot Studio vs n8n: an honest comparison

Full disclosure: our founder Marrallisa Kreijkes is an official n8n Ambassador, so we have a bias, and we have worked to keep this comparison fair.

Copilot Studio and n8n solve overlapping problems from opposite ends. Copilot Studio starts from conversational agents inside Microsoft 365. n8n starts from workflow automation across any system, with AI agents as one of its building blocks.

The n8n details come from its pricing page and licence FAQ, as of October 2026.

When Copilot Studio fits better

• You are a Microsoft 365 company and many people already have Microsoft 365 Copilot licences. Internal agents in Teams and SharePoint then add little extra cost.
• Your knowledge lives in SharePoint, Dataverse or Dynamics 365, and every answer must respect each user’s own permissions.
• IT wants one governance plane: Entra ID, Purview audit logs and Power Platform data policies for everything.
• Business teams will build and maintain simple agents themselves, without developers.

When n8n fits better

• Self-hosting and data control. You can run n8n on your own infrastructure or with a European host, so you decide where workflow data lives and which AI provider sees it. For sensitive data, that alone can settle the choice.
• Cost at volume. n8n Cloud charges per complete workflow run, however many steps it has. A self-hosted instance has no per-run licence fee at all: you pay for the server and for the AI tokens you use, directly to the provider. A high-volume process that would consume credits for every answer and action can cost considerably less this way.
• A multi-vendor stack. HubSpot, Shopify, Notion, Slack, Google Workspace, your accounting package and a custom API in one flow is exactly what n8n is built for.
• Model freedom and code. Pick a different model per step, drop into code where low-code stops, and keep workflows in Git.

n8n has honest downsides too. Self-hosting means you own updates, backups and security, or hand them to a hosting partner. It is less native inside Teams. Building it well takes technical skill. SSO and Git require paid plans. And the free Community Edition licence does not allow you to let clients build their own workflows on your instance.

If you go this route, pick a partner who designs from your processes instead of reselling templates. Here is what a real automation agency does. Already running it? Our free upgrade checklist for version 3.0 shows what to check before you update.

Or use both

Many Microsoft companies do not need to choose. A Copilot Studio agent can be the front door in Teams, while n8n does the heavy lifting across non-Microsoft systems. Copilot Studio agents can call REST APIs and MCP servers as tools, and n8n can expose a workflow through a webhook or its MCP Server Trigger. The agent asks, the workflow executes. See how the MCP Server Trigger works.

How to choose: five questions

1. Where does your data live? Mostly in Microsoft 365 points to Copilot Studio. Spread across many vendors points to n8n, or to both.
2. Who will use it? Employees with Microsoft 365 Copilot licences get included usage. Customers on your website are paid per credit.
3. How many runs a month? Model the credits with Microsoft’s estimator, then compare them with an execution-based price.
4. Where must the data be processed? If it has to stay on your own infrastructure, Copilot Studio is not an option.
5. Who maintains it after launch? Power Platform makers in-house, or a technical owner or partner?

How we help

Viralistic is a strategic digital partner in Amsterdam. We start with your process and the value you want from it, and only then choose the platform. We build automations with n8n and Claude, govern them from day one, and tell you plainly when a Microsoft tool you already pay for is the better fit. Explore our automation service, see how we govern automations and AI agents, or plan a strategic call to map your first use case.

Frequently asked questions

Is Microsoft Copilot Studio free?

Not for production use. A free trial lets you build and test agents, but you cannot publish them. People with a Microsoft 365 Copilot licence can build and use internal agents on the Copilot Chat and standard harness at no extra cost. Anything else, such as a customer agent on your website, runs on Copilot Credits: $200 for 25,000 a month, or $0.01 each on pay-as-you-go (October 2026).

Is Copilot Studio included in Microsoft 365?

Only with a Microsoft 365 Copilot licence. That licence includes Copilot Studio for internal agents used inside Microsoft 365, and classic answers, generative answers and tenant graph grounding cost licensed users nothing. Without it, your staff can still use agents, but their usage is paid in Copilot Credits. Publishing to external channels, serving unlicensed users or using the GitHub Copilot harness needs a standalone Copilot Studio plan.

What is the pay-as-you-go pricing for Copilot Studio?

It is $0.01 per Copilot Credit, as of October 2026. You link an Azure subscription to a Power Platform environment through a billing policy, and Microsoft bills the credits your agents used at the end of each month, with no upfront commitment. As a guide, a generative answer on the standard harness costs 2 credits and an agent action 5, so 1,000 generative answers cost about $20.

How do I build an agent using Copilot Studio?

Open copilotstudio.microsoft.com and describe the agent you want in plain language, and Copilot Studio drafts it. Add knowledge such as SharePoint sites, public websites or files, then add tools: connectors, agent flows, a REST API or an MCP server. Write topics where answers must be exact. Test it in the test panel, run evaluations, then publish to Teams, Microsoft 365 Copilot or your website.

Is Copilot Studio better than Copilot?

They do different jobs, so neither is better. Microsoft 365 Copilot is the assistant your employees use inside Word, Excel, Outlook and Teams, paid per user. Copilot Studio is the platform where you build your own agents and workflows, which can then appear inside Microsoft 365 Copilot, Teams or your website. A sensible order is to start with Copilot and add Studio when a process needs its own agent.

What is Copilot Studio used for?

Copilot Studio is used to build AI agents and automations for specific business processes. Microsoft’s own examples include IT support agents that handle ticket intake and triage, recruitment assistants that screen candidates, contract review agents and balance sheet reconciliation. It suits structured, repeatable work where the answers come from your own documents and systems, and where people already work in Microsoft 365.