GitHub Copilot moved into Slack and Microsoft Teams on the same day
Both shipped on August 21 in public preview. Mention @GitHub in a channel and the agent triages issues, investigates failures, writes changes in a cloud sandbox and opens a pull request, with the conversation attached. The interesting part is not the capability list, which is familiar, but the room it moved into: the place where work gets discussed rather than written.
Source
The new GitHub Copilot experience in SlackGitHub Changelog: Copilot · Original published August 21, 2026
GitHub put the same agent behind two chat clients on the same day, and the shape is identical in both. You mention @GitHub in a channel, a thread or a direct message, the agent picks up the task, and it keeps working while you go back to whatever else you were doing. Everything runs in a cloud sandbox, and the pull request it opens carries a link back to the conversation that produced it.
What it does in Slack
GitHub describes the Slack integration as bringing the agentic capabilities of Copilot CLI and the Copilot app into Slack. In practice that means four things: answering questions about your code and your GitHub activity, triaging bug reports and creating, labeling or updating issues, investigating failures and implementing the change with validation inside a secure cloud sandbox, and opening a pull request with a link to the conversation for review.
It can also create dedicated channels for a piece of work. Access requires a Copilot Business or Copilot Enterprise plan, and usage counts against the Copilot entitlements you already pay for, managed with the existing cloud agent budgets rather than a new meter.
What it does in Teams
The Teams version is framed around the shared session rather than the individual request: a discussion becomes a collaborative agent session that everyone in the channel can watch and steer. You start a task from Teams and the agent works asynchronously in the same secure cloud sandbox.
It requires a paid Copilot plan, and is also in public preview. The activation gesture is the same, a mention of @GitHub in a channel, a thread or a direct message.
The merge is still a human decision
Both announcements name the same control: repository administrators can require additional approval before a pull request authored by the agent is merged. That is the pattern every vendor has converged on, and it is worth naming each time it appears. The agent may write, run, test and propose; a person still merges.
It also tells you where the risk moved. When the agent lives in a chat client, the instruction that starts it comes from a message, and a message can be written by anyone in the channel. The approval gate on the merge is the thing standing between a badly worded request and your main branch.
Why the room matters
Nothing on either capability list is new to Copilot. What is new is that the agent stopped waiting in the editor and moved to where the decision to do the work is actually made. Bug reports arrive in chat, incidents are triaged in chat, and the gap between deciding something needs fixing and opening the tool to fix it is where tasks go missing.
Both are public preview, which is the right label for something you should try on a repository you do not mind seeing an odd pull request against.
Sources
Related
A proxy that stripped one header was doubling Claude Code's API bill
Version 2.1.239 fixes streaming on Bedrock behind proxies that remove the response Content-Type header. Claude Code silently fell back to re-running every turn without streaming, and each turn was billed twice. The same release makes cost estimates show the 1.1× premium that data-residency workspaces pay.
Anthropicverified
ChatGPT now reads and sends your Apple Messages, on the Mac only
A plug-in in the ChatGPT desktop app reads and searches your iMessage, SMS and RCS threads and sends messages through Messages on your behalf. By default nothing goes out until you approve both the text and the recipients. There is a switch that removes that step, and OpenAI's own documentation argues against using it.
OpenAIverified

Nvidia's harness takes Claude Opus 5 from about 30% to a perfect ARC-AGI-3 score
Nvidia published a run in which AVO, its agent architecture, scores 100.00 RHAE on ARC-AGI-3 and clears all 183 levels across 25 environments. The same model evaluated on its own scores about 30%. In the same post Nvidia writes that the comparison is not a controlled ablation, and that sentence did not survive into the coverage.
NVIDIAverified
