How it works

• A team member tags a supported coding agent for a feature, website update, or bug fix; Slack creates a project-specific code channel for that task.
• The channel centralizes the agent’s plan, conversation, code diffs, and a live preview, so collaborators can follow and steer the work rather than merely receiving a finished pull request.
• Team members can review diffs, provide feedback, inspect live HTML/UI previews, and approve the proposed changes before the agent’s work is merged.
• The intended governance model keeps humans accountable: teams can require human approval/sign-off before production merges. The supplied evidence does not independently document the exact configuration mechanics of that merge gate, so its precise implementation is insufficiently evidenced here.

Collaboration beyond engineering

• The point is to make coding visible to the whole product team—not only the individual developer using the agent—so product managers, designers, and other non-engineering colleagues can observe progress, comment on requirements, and evaluate the live result.
• Slack describes this as bringing software development “out of private tabs,” allowing people and agents to write, review, and ship code together in the open.

Demonstrated bug-fixing flow

Slack’s published description frames the workflow as: identify a bug (or request a new feature or webpage change), tag an agent such as Claude Tag or ChatGPT, let the agent open a code channel, then have the group inspect the conversation, diffs, and preview, give feedback, and approve the result.

Availability and agents

• Slack Code is reported as available on all Slack plans, including free workspaces.
• Launch integrations are Anthropic’s Claude Code/Claude Tag, Cognition’s Devin, GitHub Copilot, and Vercel’s agent.
• ChatGPT is described as coming soon rather than as a launch integration; thus it should not be treated as confirmed day-one availability.

Records, security, and controls

• Reports describe completed code channels as automatically archived while remaining searchable, creating a durable record of the work and decisions.
• That makes the channel an audit trail for the agent’s plan, discussions, diffs, feedback, and approval context. However, the supplied evidence does not establish the exact retention, export, or audit-log specifications.
• Slack Code is positioned to use the Slack workspace’s existing administration and security environment rather than creating a separate collaboration product. The supplied material does not give enough detail to verify precisely which security controls, permissions, or compliance tiers are inherited.

Executive framing and future scope

Marc Benioff called it: “Don’t code alone. Slack Code is live. Humans and agents. Same channel. Same work,” adding that the launch with Anthropic, GitHub, Cognition, and Vercel agents was “real multiplayer coding.”

Slack’s broader framing is that AI coding becomes a team sport: Slack is the coordination layer where the team can see, direct, review, and ship agent work together.

The idea of opening Slack APIs so agents could coordinate non-coding work—such as marketing campaigns or legal-document review—is a plausible future extension of this shared-agent-workspace model, but the supplied evidence does not verify a committed API roadmap, timing, or product scope. Insufficient evidence.

How it works

• A team member tags a supported coding agent for a feature, website update, or bug fix; Slack creates a project-specific code channel for that task.
• The channel centralizes the agent’s plan, conversation, code diffs, and a live preview, so collaborators can follow and steer the work rather than merely receiving a finished pull request.
• Team members can review diffs, provide feedback, inspect live HTML/UI previews, and approve the proposed changes before the agent’s work is merged.
• The intended governance model keeps humans accountable: teams can require human approval/sign-off before production merges. The supplied evidence does not independently document the exact configuration mechanics of that merge gate, so its precise implementation is insufficiently evidenced here.

Collaboration beyond engineering

• The point is to make coding visible to the whole product team—not only the individual developer using the agent—so product managers, designers, and other non-engineering colleagues can observe progress, comment on requirements, and evaluate the live result.
• Slack describes this as bringing software development “out of private tabs,” allowing people and agents to write, review, and ship code together in the open.

Demonstrated bug-fixing flow

Slack’s published description frames the workflow as: identify a bug (or request a new feature or webpage change), tag an agent such as Claude Tag or ChatGPT, let the agent open a code channel, then have the group inspect the conversation, diffs, and preview, give feedback, and approve the result.

Availability and agents

• Slack Code is reported as available on all Slack plans, including free workspaces.
• Launch integrations are Anthropic’s Claude Code/Claude Tag, Cognition’s Devin, GitHub Copilot, and Vercel’s agent.
• ChatGPT is described as coming soon rather than as a launch integration; thus it should not be treated as confirmed day-one availability.

Records, security, and controls

• Reports describe completed code channels as automatically archived while remaining searchable, creating a durable record of the work and decisions.
• That makes the channel an audit trail for the agent’s plan, discussions, diffs, feedback, and approval context. However, the supplied evidence does not establish the exact retention, export, or audit-log specifications.
• Slack Code is positioned to use the Slack workspace’s existing administration and security environment rather than creating a separate collaboration product. The supplied material does not give enough detail to verify precisely which security controls, permissions, or compliance tiers are inherited.

Executive framing and future scope

Marc Benioff called it: “Don’t code alone. Slack Code is live. Humans and agents. Same channel. Same work,” adding that the launch with Anthropic, GitHub, Cognition, and Vercel agents was “real multiplayer coding.”

Slack’s broader framing is that AI coding becomes a team sport: Slack is the coordination layer where the team can see, direct, review, and ship agent work together.

The idea of opening Slack APIs so agents could coordinate non-coding work—such as marketing campaigns or legal-document review—is a plausible future extension of this shared-agent-workspace model, but the supplied evidence does not verify a committed API roadmap, timing, or product scope. Insufficient evidence.