Copilot Studio adds computer-using agents and agent-to-agent communication
Diesen Beitrag auf Deutsch lesen
Copilot Studio’s May 2026 update adds generally available computer-using agents, agent-to-agent communication and a unified workflow canvas.
TL;DR
Copilot Studio’s May 2026 update makes computer-using agents and agent-to-agent communication generally available. Computer-using agents automate user interfaces without an API, while the visual designer supports multi-agent orchestration without an SDK. A redesigned workflow canvas combines deterministic, branching and AI steps, with the source reporting approximately 20% better evaluations and 50% lower token usage.
Original by Divya Akula, on Divya Akula. Read the original
This is our own summary, not a republication or full translation.
Why it matters
- Makers: Test computer-using agents against portal tasks that previously lacked APIs, because the feature supports UI automation with enterprise credentials and adaptive interface recovery.
- Admins/IT: Review Copilot Studio environments for the May release and assess agent-to-agent communication, because Studio-built and SDK-built agents are described as wire-compatible.
- Leadership/Business: Recalculate running-agent costs under Copilot Credits, because consumption pricing became effective on June 1 and pay-as-you-go is now available.
Frequently asked questions
How do Copilot Studio computer-using agents automate portals without APIs?
Copilot Studio computer-using agents automate through the user interface, so an API is not required. Enterprise credentials and adaptive interface recovery are built in.
What does Copilot Studio agent-to-agent communication enable?
Agent-to-agent communication enables multi-agent orchestration through a visual designer without an SDK. Studio-built and SDK-built agents are described as wire-compatible.
How does the Copilot Studio unified workflow canvas change automation design?
The unified canvas combines deterministic, branching and AI steps in one workflow. The source reports approximately 20% better evaluations and 50% lower token usage.