Craig Stanley
Home / Capabilities / Copilot Studio / Reading a Copilot Studio bill in Copilot Credits

Reading a Copilot Studio bill in Copilot Credits

How to turn an agent's design into a Copilot Credits estimate, using Microsoft's published rates and a worked example you can copy.

· 3 min read · Craig Stanley
In short, explained

Every time an agent does something, it uses up some credits, like coins in a meter. Different jobs cost different numbers of coins.

Copilot Studio charges in Copilot Credits. Each thing an agent does has a price in credits: a written-out answer, a search of company data, an action. Add up what one conversation uses, multiply by how many conversations you expect, and you have an estimate.

Credits = Σ(feature events × rate) per run × runs. Rates: classic 1, generative 2, tenant graph grounding 10, agent action 5, flow actions 13/100, AI tools 1/15/100 per 10 responses, reasoning premium 10 per 1K tokens. Licensed employee-facing use under user identity is no charge. Monitor in PPAC; enforce per-agent monthly caps.

The rate card

Microsoft publishes the rates on its Copilot Studio billing page. These are the ones I use most:

FeatureCopilot Credits
Classic answer (a fixed, authored reply)1
Generative answer2
Agent action5
Tenant graph grounding for a message10
Agent flow actions13 per 100 actions
Text and generative AI tools, basic1 per 10 responses
Text and generative AI tools, standard15 per 10 responses
Text and generative AI tools, premium100 per 10 responses

When an agent uses a reasoning model, Microsoft adds a premium charge of 10 credits per 1,000 tokens on top of the feature rate.

Pay-as-you-go credits are $0.01 each. Prepaid packs give 25,000 credits a month.

The rule that saves the most money

Microsoft says employee-facing use of Copilot Studio agents is included in the Microsoft Copilot licence when the user is licensed and the agent runs under that user's identity. For agent flows, that only applies to flows triggered by "When an agent calls the flow". Anything triggered another way, such as on a schedule or a new record, is charged.

So the same agent can be free for one person and charged for another, and an autonomous agent is charged whoever benefits from it.

How to read one conversation

Microsoft's example is a good template. A tenant-grounded answer to one complex prompt uses 12 credits: 10 for grounding and 2 for the generative answer. To estimate an agent, I write down one typical conversation as a list of events, give each a rate, and add them up.

A worked example

The figures are illustrative. A procurement agent helps staff decide whether a purchase needs a formal tender. A typical conversation has:

EventCountRateCredits
Generative answers326
Tenant graph grounding21020
Agent action (log the decision to a list)155
Total per conversation31

If 600 conversations a month come from people without a Copilot licence, that's 18,600 credits, or $186 at pay-as-you-go rates. Conversations from licensed staff under their own identity add nothing. If the agent switched to a reasoning model using about 2,000 tokens a conversation, each would add roughly 20 credits, taking the unlicensed total to about 30,600 credits.

That last line is the one I'd watch. Model choice can change the bill more than usage does.

Where to see the real numbers

Microsoft says consumption is reported in the Power Platform admin center. For agent flows, it points to Licensing, then Copilot Studio, then Environments, where the Agent flow actions line shows credit use. Admins can set a monthly credit limit for each agent under Licensing, Copilot Studio, Manage Agents. Microsoft also links an agent usage estimator from the billing page, which I'd use alongside a hand count like the one above.

What I'm still checking

I haven't found a simple way to see credits per conversation for a specific agent, which is the number I'd want to compare with the estimate. The environment-level view tells you the total. I'd run a short pilot with known conversation counts to back out the real figure.

Sources

Read next

A question to take awayWhich of these do you already pay for and not use?

About me

Craig Stanley

Microsoft AI consultant and technical architect, based in Whitley Bay. Over the last few years I've delivered Microsoft 365 Copilot, Copilot Studio agents, Microsoft Foundry (formerly Azure AI Foundry) work and governance for UK public sector and financial services organisations.

What interests me is the decision underneath the tool: what it costs, what it risks, and whether a small, transparent model can make it better. I write the methods up here and on Substack so anyone can use them.

I write this site to learn in public: explaining each idea simply is how I check I understand it. Why I write this site.

Find me