Slack Launches Code Channels So Teams Can Ship Software Without Leaving Slack

Slack launched Slack Code on August 20, 2026. It’s a new channel type that keeps AI coding agents inside the conversations where your team already works, rather than asking everyone to log into a separate tool to review what the agent did.

The basic mechanic is straightforward: mention an agent in a Slack channel or DM, and a dedicated code channel spins up for that project. Everyone on the team gets the same view the agent is working from. They can watch code diffs appear in real time, see live HTML previews of the output, leave feedback the agent incorporates, and approve the finished work before it ships. When the project closes, the channel archives automatically. For anything that touches production, a channel expert must approve the action before it executes.

Slack says over 70% of code channels close within a single day, conception to merged pull request.

The Launch Partners

Four agents are available now:

Claude Code (Anthropic)
Devin (Cognition)
GitHub Copilot (GitHub/Microsoft)
Vercel Agent

OpenAI’s ChatGPT integration is listed as coming soon.

GitHub is going live as a partner here as part of a related announcement: the new GitHub Copilot experience in Slack. Mention @GitHub in any channel or DM to start a Copilot session, ask questions about your code and GitHub activity, triage or create issues, investigate failures, and hand off implementation work. GitHub is treating Slack Code as a first-class surface rather than an integration bolted on after the fact. Their Chief Product Officer Mario Rodriguez put it plainly: “Humans set direction, agents close the loop. That has to be true wherever developers are working.”

Claude Code in Slack works the same way it works in the terminal or IDE: the agent can access repositories, run code in a secure sandbox, and push changes as pull requests. The Slack channel just gives the rest of the team a window into that process instead of the work happening invisibly in someone’s local session.

Why This Matters

The pattern Slack is establishing here is worth paying attention to. Most AI coding workflows today are solo: one developer, one agent, one editor. The bottleneck is still the human who has to review what the agent built and decide whether it’s shippable.

Slack Code makes the review process a group activity inside the tool where teams already communicate. The audit trail, diff reviews, and approval gates are built in. The channel archives when you’re done. There’s no separate dashboard to check or another product to learn.

Availability is immediate for teams using the participating agent integrations. Slack Code works on any Slack plan.

Sources: