Microsoft Copilot Studio Platform - Build 2026.6.1

This article describes the fixes and improvements to existing functionality that are included in this build.

Visit the Microsoft Copilot Studio Blog to stay up-to-date on the latest news!

Version 2026.6.1 updates

New features

• Custom metrics analytics now support application lifecycle management so makers can manage and move custom metrics configuration across environments more easily.
• Streaming connector calls can be switched to use service-to-service authentication instead of on-behalf-of authentication to reduce failures when access permissions are missing.
• Maker evaluation can include content safety grading across hate, self-harm, sexual, and violence categories with configurable severity thresholds.
• Workflow evaluations can persist customer-provided properties sent when starting a run, and return those properties with run results, including the flow trace identifier and start time.
• Dataverse search capability is enabled by default.
• Makers can override an agent’s prompt text using environment configuration settings, enabling prompt updates without redeploying the agent.

Fixes and improvements

• Conversation handling is more robust in long chats. Agents using Anthropic models and GPT-5 planners more consistently stay within context window limits by trimming older history while protecting the active turn, and agents using Anthropic models are less likely to fail in SubOrchestration or reject requests when multiple planner calls occur in the same turn.
• Authoring, publishing, and provisioning are more predictable. Sharing or unsharing a saved but unpublished agent completes successfully, publishing is no longer blocked by Domain Language Model training being off, enhanced task completion enabled agents provision the required agent user during publish or fail clearly, and agent registry responses consistently include fields such as SharedWith, Editors, language model settings, server connectors, and triggers.
• Analytics experiences are clearer and more accurate. Adaptive Card content and submitted values display more reliably in transcripts, Questions exploration shows the newest questions first with correct totals, Theme details show KPI tiles for answered versus unanswered questions and reactions, and custom metrics analytics include sessions with only agent messages for more consistent results.
• Error messages and status codes are clearer. Strict mode without an activity template returns a validation error, Conditional Access blocks return 403 instead of 500, temporary Microsoft Graph issues return 503 instead of 500, runtime errors no longer start with “Knowledge failed,” and empty model responses are labeled clearly.
• Chat and tool integration behave more reliably. Agents preserve the user’s last message when a button click sends an empty message, agents that use flow actions emit the subscribe trace payload by default so flow completion is signaled back to chat, chats show flow completion for async subscribe-endpoint channels, and connected agents in Microsoft Teams channels respond to the first user message more reliably.
• Continued improvements to performance and reliability.
• Breaking changes: Agents now return an error when the agent definition cannot be retrieved, instead of returning a null result.

Security

• Agents that use restrictive access settings for SharePoint are accessible only to the intended users and groups, preventing broader site readers from gaining access.
• Publishing an autopilot (agent) can be blocked in environments where agent user provisioning is disabled by an admin governance setting, while already provisioned autopilots remain unaffected.

Return to the Microsoft Copilot Studio Release Notes.