Microsoft Copilot Studio is Microsoft's low-code environment for designing, testing and publishing AI agents. It sits inside the Power Platform and Microsoft 365 ecosystem, letting makers wire agents to enterprise data, Teams, Dynamics 365 and third-party APIs without heavy custom code.
The current version is 2026 release wave 1, running on platform build 2026.6.2 (June 2026). This wave brought several substantial changes: computer-using agents reached general availability, and a new orchestration layer improves accuracy on evaluations by roughly 20% while cutting token consumption by about 50%. Real-time speech-to-speech voice agents were added for Dynamics 365 Contact Center, and multi-step workflows were redesigned so UI-driven agents run alongside API actions and approvals.
On the capability side, the platform now supports multi-agent orchestration, autonomous computer-using agents and a broader model choice. In the Prompt Tool you can select GPT-5.5 Reasoning (early release) or Anthropic Claude Opus 4.6 and Sonnet 4.5, offered as a paid experimental preview in the US.
Pricing runs entirely on Copilot Credits, which replaced the old per-message billing on September 1, 2025 (same quantity and rate). A prepaid capacity pack is $200 per pack per month for 25,000 Copilot Credits; pay-as-you-go is metered at $0.01 per credit with no upfront commitment but requires an Azure subscription. A one-year Copilot Credit Commit Units (CCCU) prepurchase can save up to 20%. Building agents is included at no extra cost with a Microsoft 365 Copilot license at $30 per user per month, billed annually.
The main limitation is that there is no free production tier. The free authoring or maker license costs $0 but only works if the tenant already holds a paid prepaid Copilot Credit pack, and the individual free trial license is time-limited and cannot publish agents. In practice, deploying anything requires payment.
Copilot Studio suits organizations already invested in Microsoft 365, Dynamics 365 or Azure that want governed, low-code agents connected to their existing data. It is a weaker fit for individuals or small teams wanting a genuinely free way to ship an agent to production.
Leave a Review
Reviews are published after moderation. We don't share your email.