GitHub Copilot is expanding beyond the IDE.
GitHub added the option to open shared Copilot sessions from Slack and Microsoft Teams.
A team can mention @GitHub, begin investigating a problem, plan work and follow a task the agent performs.
What GitHub added
The idea is to turn an existing development team conversation into the starting point for an AI task.
Instead of:
- Reading the discussion
- Copying requirements
- Opening an IDE
- Writing a new prompt
The workflow can begin within the conversation itself.
Other team members can see the session and continue directing it.
Why this is a significant change
Coding agents are gradually turning from a tool serving one user into a team tool.
Many development problems begin somewhere else entirely:
- Slack
- Teams
- Jira
- Azure DevOps
- A monitoring system
- A customer ticket
GitHub wants to connect that starting point to the agent's task.
Copilot App also received a Customize area
In the same update wave, GitHub made Copilot app's Customize area generally available.
The area brings together:
- MCP servers
- Plugins
- Skills
- Canvases
This means Copilot is becoming more of a composable system and less of a product with a fixed capability set.
Azure DevOps connects too
Issues and Pull Requests from Azure DevOps can serve as starting points for a Copilot session.
This matters particularly for organizations using Microsoft and GitHub simultaneously.
What changed in the CLI
GitHub added to Copilot CLI:
- Saved defaults for permission and execution modes
- Plugin management
- MCP server management
- Skill management
- Recovery of interrupted sessions
- A new runtime in Rust
GitHub says the new runtime is faster.
This is GitHub's claim, not independently tested.
Why this matters for businesses
When a coding agent operates from Slack or Teams, it is no longer just an individual developer's tool.
You also need to think about:
- Who may activate it
- Which repository it can access
- Who can approve actions
- How an audit trail is retained
- How cost is measured
- And what happens when an external participant is in the conversation.
What has not yet been published
GitHub did not publish a full breakdown of the plans supporting each integration in its weekly summary.
Regional availability is not detailed either.
The exact AI credit cost of a shared session is unclear.
No explicit commitment to Hebrew support in these sessions was found.