Slack Code Brings AI Agents to Shared Workspaces

Slack launched Slack Code on August 20, 2026, creating dedicated channels for teams to build software alongside AI coding agents. [2] Instead of individual developers working with agents in private, teams can tag an agent in a conversation, prompting it to open a project-specific channel. [2] In these channels, participants can review proposed changes, see live HTML previews, and provide feedback. [2][4]

These code channels function as the unit of work, automatically archiving themselves when an assignment is completed. [2] They maintain an audit log detailing who requested the task, the changes made by the agent, and who approved them. [2] Slack asserts that anyone in a code channel can pause or stop an agent, and high-stakes actions like pushing code require expert approval. [4]

The feature is available on all Slack plans, including free workspaces, immediately upon launch. [2] Slack Code integrates with agents from the Slack marketplace, including Anthropic's Claude Code, Cognition's Devin, Vercel Agent, and GitHub Copilot. [2]

Slack promotes this approach as a way to include non-engineering staff, such as product managers and designers, in the development process without requiring them to use specialist tools. [4] Agents operating in Slack Code inherit the platform's existing permissions and security policies, removing the need for a separate management layer. [4]

Slack Code represents a shift from solitary AI-assisted coding in IDEs to a collaborative, platform-based model. By embedding agents directly into the communication hub, Slack aims to streamline workflows and increase visibility across disciplines, much like a shared Google Doc does for writing. The integration of established agents like Claude Code and GitHub Copilot signals strong industry backing for this "multiplayer AI" approach. What the sources don't address: how organizations will manage the potential noise and access control implications of widespread agent-driven channels.