Slack adds web-coding to chat with dedicated code channels

Slack announced today that it is rolling out Slack Code, a new layer that turns ordinary conversation channels into collaborative coding workspaces where teams can write code alongside AI agents without switching tools. Instead of hopping between an IDE, a terminal, a GitHub pull-request view and a separate chat, an agent—such as Claude Code from Anthropic, Devin from Cognition, Vercel Agent or GitHub Copilot—creates a dedicated code channel for the task. In that channel all participants see the ongoing dialogue, real-time diffs, a preview of HTML output, and can approve changes before the code is promoted to production.

According to the company’s announcement, when a feature idea, bug fix, or page update is raised, the relevant agent is tagged in the existing channel. The agent then opens a separate code channel for that specific task, makes the changes, and presents them as a diff against the current codebase. Once the task is completed, the channel archives itself automatically and an audit log is retained for documentation. The feature is available today across all Slack plans and works with the agents already listed in the platform’s marketplace; Slack says the four tools “integrate seamlessly” with code channels.

The underlying model is as much organizational as technical. Slack treats agents as “team members” that receive a channel, participate in the conversation, expose their reasoning, and wait for human approval before merging. This mirrors the approach SpaceXAI is pushing with its Grok Bot agents: turning AI from an external “code-throwing over the wall” service into an integral part of the daily flow, with full transparency and a retrievable history.

The announcement does not detail the API-level integration, the permissions required for an agent to access code repositories, or any benchmarks comparing output quality or speed against a traditional IDE equipped with Copilot or Cursor. It also leaves unanswered how conflicts are resolved when two or more agents work on the same component simultaneously, or what the review experience looks like for large diffs. Slack notes only that archiving and the audit log are automatic; policies on data retention, encryption, or regulatory compliance were not disclosed.

Slack Code is an attempt to close the gap between the place where code is discussed and the place where it is written and reviewed. If the integration with agents truly “integrates seamlessly” as the company claims, and if the audit log meets the requirements of security and compliance teams, the feature could become the default for Slack-centric development teams. Without independent data or field testing, however, it remains a product promise rather than a proven reality.