On this page
Put Coworker to work on your stack.
Connect Salesforce, Slack, Jira and run your first agent in minutes.
Book a demoFree for 14 days
Enterprise AI
Copilot Studio Pricing in 2026: What Credits Actually Cost
Coworker AI breaks down Copilot Studio pricing in 2026: the free maker license, $0.01 per Copilot Credit, $200 capacity packs, and what drives the real bill.
How much does Copilot Studio cost?
Copilot Studio pricing has no per-user fee. The maker license is free of charge, and billing runs entirely on Copilot Credits, which cost $0.01 each on the pay-as-you-go meter or $0.008 each if you prepay through a capacity pack (Azure pricing, Microsoft Learn).
That single design choice is why Copilot Studio budgets are so hard to predict. There is no headcount number to multiply. What you are really buying is agent activity, and the same agent can cost wildly different amounts depending on how it is built.
| Buying option | Cost | How it works |
|---|---|---|
| Maker license | Free | Assigned in the Microsoft Admin Center. Required to build. |
| Pay-as-you-go | $0.01 per Copilot Credit | Billed monthly through an Azure subscription. No commitment. |
| Capacity pack | $200/month for 25,000 credits | Roughly $0.008 per credit. Pooled across every agent in the tenant. Packs stack. |
| Pre-purchase (CCCU) | Custom, up to 20% off | Prepaid Copilot Credit Commit Units bought annually, with automatic pay-as-you-go overflow. |
What is a Copilot Credit?
A Copilot Credit is the billing unit for one interaction with an agent. Until recently it was called a message. Microsoft changed the currency from messages to Copilot Credits on September 1, 2025, with no change to the quantity per prepaid pack or to the pay-as-you-go rate (Microsoft Learn).
If you are reading a pricing guide that still talks in messages, it predates that change. The economics are the same, but the terminology in your invoice will not match.
Why the same agent can cost 5x more than another
Not every interaction draws the same number of credits. Microsoft's rate card bills different capabilities at different rates, and the spread is wide:
- Scripted topics draw the least. A deterministic flow answering a known question is the cheap case.
- Generative answers draw more, because the model is retrieving and composing rather than following a script.
- Autonomous actions draw the most, because the agent is deciding and acting on its own.
Rates also exclude bring-your-own-model configurations. If you point an agent at an Azure Foundry model, that inference is billed separately on top of your credits (Microsoft Learn).
The practical consequence: a handful of popular generative agents can outspend a large fleet of scripted ones. Estimating from a demo is how teams get surprised, because a demo is almost always the cheap path.
Pay-as-you-go or capacity packs?
The arithmetic is simple. A capacity pack is $200 for 25,000 credits, which is $0.008 per credit against $0.01 on pay-as-you-go, so prepaying saves about 20% on the credits you actually use.
That discount only pays off if you use the pack. Buy 25,000 credits and burn 9,000 and your effective rate is $0.022 per credit, worse than pay-as-you-go. The honest rule is to start on the meter, measure a full month of production traffic, then convert to packs once your floor is genuinely predictable.
Capacity is pooled across the tenant rather than locked to one agent, which helps. Packs also stack, so you can size up without renegotiating.
Coworker
See what your AI stack really costs
Compare model and platform costs, then run it all in one place.
Open the free LLM cost calculatorWhat the sticker price leaves out
Three costs sit outside the credit meter and belong in any real estimate.
Prerequisite licensing. Copilot Studio sits in the Power Platform. Agents that touch Dataverse, premium connectors, or governed environments pull in licensing and administration that is not on the Copilot Studio price list.
Governance overhead. Credits are pooled tenant-wide, which is convenient until one team's experiment drains the shared pool. Someone has to own allocation, monitoring, and shutting off runaway agents.
Estimation drift. Independent consultancies reviewing live deployments report production consumption running meaningfully above initial estimates, because early sizing is modelled on scripted demos rather than the generative traffic that shows up in production (Redress Compliance). Treat that as a directional warning from practitioners rather than a published Microsoft figure, and build headroom into the first quarter.
How Copilot Studio pricing compares
Consumption pricing is not unique to Microsoft. It is close to the norm across agent platforms, which makes cross-vendor comparison harder than it looks.
| Platform | Pricing shape | Entry point |
|---|---|---|
| Coworker AI | Per seat | Pro $29.99/user/month |
| Copilot Studio | Consumption (credits) | $0.01 per credit, $200 capacity packs |
| Microsoft 365 Copilot | Per seat | See our Microsoft Copilot enterprise pricing breakdown |
| Gumloop | Subscription plus usage | $37/month plus an 8% orchestration fee |
| Salesforce Agentforce | Consumption | See Agentforce pricing |
Per-seat and per-credit pricing fail in opposite directions. Seat pricing punishes you for adding people who use the tool lightly. Credit pricing punishes you for success, because the better your agents work, the more they run. Neither is inherently cheaper. The question is which failure mode your organisation can forecast, and consumption models are considerably harder to forecast in year one.
If you want the broader market view, our enterprise AI pricing comparison puts the major platforms side by side, and how much enterprise AI should cost per user works through the per-head maths.
Where Coworker AI fits
Coworker AI prices per seat: Pro at $29.99 per user per month, Max at $149.99, and Enterprise on a custom plan. You know the bill before the month starts, and an agent that runs more does not cost more.
Coworker connects to 50+ enterprise tools and includes an agent builder, so the comparison with Copilot Studio is genuinely like for like on capability. The difference is the billing model and the breadth of what it connects to outside the Microsoft estate. If your stack is entirely Microsoft, Copilot Studio's integration depth is a real advantage and worth weighing honestly. If it spans Slack, Salesforce, Jira, Zendesk, Google Workspace and the rest, a platform that treats those as first-class is usually the better fit.
Related depth: AI agent orchestration platforms, best enterprise AI workflow automation, and our Microsoft Copilot alternative comparison.
Book a demo to see how the numbers work against your actual usage.
Frequently asked questions
Is Copilot Studio free?
The maker license is free of charge, so building costs nothing. Running agents is not free. Every interaction draws Copilot Credits billed at $0.01 each on pay-as-you-go, or $0.008 through a prepaid capacity pack.
How much is a Copilot Studio capacity pack?
$200 per month for 25,000 Copilot Credits, which works out to about $0.008 per credit. Packs are pooled across every agent in the tenant and can be stacked.
What replaced messages in Copilot Studio pricing?
Copilot Credits. Microsoft changed the billing currency from messages to Copilot Credits on September 1, 2025. The quantity per prepaid pack and the pay-as-you-go rate did not change.
Does Copilot Studio charge per user?
No. There is no per-seat fee for Copilot Studio itself. Billing is consumption-based on Copilot Credits, so cost tracks agent activity rather than headcount.
Why is my Copilot Studio bill higher than expected?
Usually because the mix of interactions shifted. Scripted topics draw the fewest credits, generative answers draw more, and autonomous actions draw the most. An estimate built on a scripted demo will understate a production workload that leans generative.
Is pay-as-you-go or prepaid cheaper for Copilot Studio?
Prepaid capacity packs are about 20% cheaper per credit, but only if you consume the full pack. Underuse a $200 pack and your effective rate rises above the $0.01 pay-as-you-go rate. Measure a full month on the meter before committing.
Related reading
Ready to get started?
Put Coworker to work inside your actual stack
Connect Salesforce, Slack, Jira, whatever you use, and run your first agent in minutes.
Free for 14 days