As of June 16, 2026, Work IQ MCP servers are billed using Copilot Credits — and any Copilot Studio agent that uses them is charged per API call. The headline figure circulating is small: about 0.1 Copilot Credit, or roughly 0.001 USD, per call. The catch is the one every engineer spots immediately: MCP servers rarely make a single call. One agent turn fans out into many tool calls, so the real question is not the per-call price — it is what a whole task costs, and whether you can forecast it.
Here is what is actually confirmed, what is still a wide range, and how to keep it under control.
What changed on June 16, 2026
Per Microsoft's Work IQ MCP overview, from June 16, 2026 Work IQ is available through consumption-based billing using Copilot Credits — the common currency across Copilot Studio capabilities and Work IQ protocols in your tenant. You still need a Microsoft 365 Copilot license as the entry point, and you pay for usage on top through pay-as-you-go or a Copilot Credit prepurchase (P3) plan.
This is the same metered model Microsoft just shipped for Copilot Cowork: the license gets you in the door, the agent work is billed by consumption.
What Work IQ MCP actually is
Work IQ is Microsoft's workplace-intelligence layer — Data, Memory, and Inference — that grounds Copilot and agents in real-time organizational context. Work IQ MCP tools (Work IQ Mail, Work IQ Calendar, Work IQ Teams, plus custom servers) let your Copilot Studio agents reason over emails, meetings, documents, Teams messages, people, and enterprise search results — permission-trimmed, with Microsoft Entra delegated authentication and compliance policies enforced automatically.
If the term is new to you, it is worth understanding what an MCP server is first: it exposes systems as tools an AI can call. Work IQ MCP is Microsoft's managed set of those tools for Microsoft 365 data.
How does the billing work?
Copilot Credits are priced at 0.01 USD per credit. The figure Microsoft and the community have attached to Work IQ MCP is roughly 0.1 credit — about 0.001 USD — per API call. On its own that sounds trivial.
A note on precision: Microsoft's published overviews confirm the Copilot-Credit billing model, the June 16 start, and the cost controls below, but they do not publish a single fixed per-call rate on those pages. Treat the 0.001 USD/call number as the reported rate, and pay attention to the multiplier instead — because that is where the money is.
What does a task actually cost?
Microsoft has shared a few consumption scenarios, and the ranges are wide enough to make the point. These are per task, not per call:
- Light — "identify tasks or action items assigned to me from my manager and compile them into a checklist": roughly 0.20–0.40 USD.
- Medium — "review the latest customer interview emails to identify top themes and roadmap impact, and recommend three prioritized product actions with alignment, tradeoffs, and objections": roughly 0.30–0.75 USD.
- Heavy — "produce Level 1 and Level 2 summaries from our most recent roadmap executive review, using the latest meetings, documents, and what is top of mind for our CMO": roughly 0.50–1.50 USD.
A spread of 0.20 to 1.50 USD per task is not a price; it is a planning problem. Multiply by the number of users and how often they run these tasks, and the monthly figure becomes genuinely hard to predict at budgeting time — the same total-cost-of-ownership uncertainty that comes with any metered agentic service.
Where do the charges show up?
In the Cost Management dashboard in the Microsoft 365 admin center. Its Overview tab gives a real-time snapshot of credit consumption and remaining capacity; the Consumption tab lets you drill into usage by user, group, service, or agent to find your cost drivers. That dashboard is also where Copilot Studio consumption is set to appear, so it is the one screen to watch.
How do you control or switch it off?
You are not at the mercy of the meter:
- Block it outright. Admins allow or block Work IQ MCP servers in the Microsoft 365 admin center under Agents and Tools. Blocking a server blocks it for every user and every agent, and permissions always take precedence over configuration. If you do not want to be billed for Work IQ MCP, this is the off switch.
- Cap the spend. The Cost Management dashboard supports spending policies, budgets, alerts, and hard caps across pay-as-you-go and prepaid credits.
- See what ran. Observability is built in through Microsoft Defender Advanced Hunting — you can trace which tools were invoked, with which parameters, and spot anomalous usage.
Takeaways for IT decision-makers
- Watch the Cost Management dashboard. That is where these charges land today, and soon Copilot Studio consumption as well.
- Justify the spend. In this metered era of AI, every agent run has a price — tie usage to high-value scenarios rather than switching everything on.
- Turn off what you do not need. If Work IQ MCP is not earning its keep, block it via admin and data policies before it quietly accrues credits.
- Keep deterministic work deterministic. For repeatable, predictable processes, purpose-built automation on Azure and Microsoft Graph is cheaper and more forecastable than paying metered per-call runtime. Reserve Work IQ MCP for the judgment-heavy, cross-system tasks where grounding in live context genuinely pays off.
Metered, per-call AI is the direction Microsoft is taking across the Copilot line. It is fair — agentic work consumes real compute — but it moves the cost risk onto you. Turn it on deliberately, cap it, and measure it.