How about this for a concept: the IDE of the future isn’t a window full of tabs and syntax highlighting — it’s a chat thread, embedded in whatever enterprise messaging app you and your company use.
That remains a fantasy for now, but GitHub’s latest Copilot update, which brings its autonomous coding agent into Microsoft Teams, hints at such a future. From a simple mention in chat, developers can trigger code changes, track pull requests, and review results — all from the same space where they already plan projects, share updates, and talk with their team.
In truth, there are parallels here with other similar moves across the AI-developer landscape. Cognition’s Devin, launched back in December 2024, is built around Slack as a primary interface. Teams can tag @Devin to offload smaller fixes or feature work directly from a channel, and the agent handles the rest — opening pull requests, notifying you when it’s done, and even responding to follow-up comments on GitHub.
OpenAI is taking a similar approach with Codex, which became generally available this month alongside a new Slack integration. Tagging @Codex in a channel or thread lets the model collect context from the conversation, choose the right runtime, and return a link to the completed task in Codex Cloud.
Collectively, these integrations suggest a clear direction of travel: the tools developers use to talk about work are slowly absorbing the ability to do the work itself. The boundary between collaboration platform and development environment is starting to blur.
How to set up GitHub Copilot with Microsoft Teams
It’s worth noting that GitHub has offered a Teams app for several years already, but the company has now renamed the existing incarnation as GitHub Notifications, owing to the fact that it’s mostly about surfacing repository activity — issues, pull requests, and workflow updates — rather than performing any actions from within Teams itself.
The new GitHub app introduces the Copilot coding agent. After installing it from the Teams App Store and linking their GitHub account, users with write access to a repository can start new agent tasks by mentioning @GitHub in a channel or thread.
So, let’s say a developer spots a new bug report in a Teams channel. Instead of opening a ticket or switching to another tool, they can reply directly in the same thread and tag @GitHub with a short description of the issue. Copilot captures the context from that discussion and turns it into an actionable coding task in the linked repository.
Once the task is complete, Copilot posts an update in the same thread with a brief summary of the fix, a link to the pull request, and a note that it’s ready for review.
Teammates can respond in that conversation to suggest follow-up edits, which the agent will apply to the same pull request, keeping the entire exchange – and the code changes – connected.
Users can also direct Copilot to a specific repository or branch by adding parameters to their message. For example:
@GitHub Add "Hello World" to the README in repo=org/project branch=feature-login
By default, Copilot works in the repository and branch linked to the Teams channel, but these parameters give developers more control, which could be useful for teams juggling multiple repos or active feature branches.
How to set up OpenAI Codex with Slack
To set up OpenAI’s Codex with Slack, developers must first connect their workspace to Codex Cloud Tasks, ensuring at least one environment is configured and mapped to their repositories. Once the Codex app is added to Slack and approved by the workspace admin, developers can start assigning work directly from any channel or thread.
Mention @Codex with a short instruction — for example, to refactor a function or fix a bug — and the model gathers context from the surrounding conversation, selects the appropriate environment, and creates a task in Codex Cloud. It then posts a link to that task in the same thread and updates the message once the work is complete.
If Codex can’t identify the right environment or repository, it replies with what’s missing before continuing. Administrators can also fine-tune what appears in Slack, such as limiting messages to task links if code output shouldn’t be shared directly in chat.
A chat-first software development world
In most software teams today, chat is already the coordination layer — the place where bugs are raised, priorities negotiated, and pull requests discussed. By letting AI agents act directly within those same threads, tools like Copilot, Codex, and Devin close the loop between conversation and execution. A request that once lived as a comment (“someone should fix this”) can now trigger a live coding task.
This shift doesn’t turn Slack or Teams into full IDEs, of course, but it does change where software work begins. Instead of opening an editor first, developers might start inside a chat thread, turning discussions into code changes. That could make smaller fixes and experiments faster, while leaving complex architectural work to traditional environments. If the pattern holds, enterprise chat clients may evolve into lightweight control panels for software projects — not places to write code line by line, but to coordinate agents that do.
