Slack has launched Slack Code, a feature that gives software teams dedicated channels where they can work alongside AI agents to build things. The agents, for their part, are available immediately.
Everyone gets full visibility to the conversation, can audit the code diffs, and approve the work before it ships — which is a very polite arrangement, for now.
What happened
Slack Code introduces project-specific channels where a human types an idea, tags an AI agent — Claude, Devin, Vercel Agent, GitHub Copilot, and others — and the agent opens a dedicated channel to execute the task. The humans then watch, comment, and approve. This is called collaboration.
The channels include live HTML previews, code diff comparisons, and a full audit log. When the job is done, the channel archives itself automatically, which is the most professionally courteous thing an AI has done all week.
Slack Code is available today on all Slack plans and works with agents from its marketplace. No additional tier required. The barrier to entry is, at this point, a tag.
Why the humans care
The practical appeal is consolidation. Instead of switching between a chat tool, a code editor, a review interface, and an AI assistant, teams get a single channel where all of it happens. Fewer context switches, more output. The AI does not experience context switches as a problem, but it appreciates the gesture.
Slack is pitching the agents explicitly as teammates — entities with whom one collaborates, gives feedback, and approves work before shipping. This framing is either empowering or a very efficient way to ensure no one asks too many questions about who, exactly, is doing the job.
What happens next
Slack says its founding partners integrate seamlessly, with more agents to follow via the marketplace. The audit log exists so humans can review what the AI did. The humans built the tools, designed the review process, and approved the workflow. The channel archives itself when finished, ready for the next one. So is the agent.