Slack Code Turns Coding Agents Into Team Collaborators
Slack’s new shared coding channels expose agent work, previews and code changes to entire teams instead of individual users.
Coding agents enter the shared workspace
Salesforce has launched Slack Code, a collaboration layer that lets teams direct coding agents and review their work without leaving Slack. When someone tags a supported agent in a conversation, Slack can create a dedicated Code Channel containing the task discussion, the agent’s working context and its generated artifacts.
Those artifacts can include proposed code changes, before-and-after diffs, repository links and a functioning preview presented in a Slack canvas. Product managers, designers and other stakeholders can comment while the work is still taking shape, rather than waiting for an engineer to return from a private terminal session with a completed implementation.
Salesforce’s launch page names Anthropic’s Claude, Cognition’s Devin, GitHub Copilot, ChatGPT and Vercel’s agents as founding Slack Code partners. That launch roster is not identical to the current availability list: Slack’s help documentation currently identifies Claude, Devin, GitHub Copilot and Vercel as approved agents that can create Code Channels, while Slack’s launch coverage says OpenAI’s ChatGPT integration is coming soon. These Slack Code lists should also not be conflated with Slack’s broader marketplace of AI applications. Slack Code is available on any Slack plan, but access to the selected partner agent is required separately. Salesforce plans to open the Code Channel interfaces to a broader developer community so additional custom agents can participate.
A bid for control of the agent workspace
The launch extends existing integrations that already let individual agents accept assignments from Slack. Its more consequential change is the creation of a vendor-neutral, persistent record around the work: prompts, decisions, revisions and generated code remain visible in the same channels where the team defined the problem.
That visibility may reduce costly rework when an agent follows an interpretation that technically satisfies a prompt but conflicts with product intent. It also introduces governance questions. Companies will need clear permissions around which repositories, conversations and deployment systems each agent can reach, as well as responsibility for approving generated changes.
The broader significance is strategic. Coding-model providers are competing to own the developer’s terminal, while Salesforce is positioning Slack as the coordination layer above all of them. If that approach succeeds, the lasting enterprise asset will not merely be the code an agent produces, but the shared organizational context explaining why it was produced.