Slack announced the launch of Slack Code, a feature that places AI coding agents inside dedicated channels on the messaging platform, rather than limiting a developer’s interaction with an agent to a private terminal window. At launch, the feature supports agents including Anthropic Claude Code, Cognition Devin, GitHub Copilot, and the Vercel agent.
Slack Code allows team members to view, direct, and review a coding task together, displaying code diffs, live previews, and the ongoing work plan in dedicated tabs. When the task is complete, the channel is archived with a searchable record, providing a reference for what took place during its execution. The feature is available on any Slack plan, but using partner agents requires the customer to have access to them.
How Do Coding Channels Work?
Users can mention a coding agent from any conversation, prompting Slack to create a channel linked to the project and task. In a demonstration by Cognition, a member of the engineering team reported a bug in a dedicated channel, and Devin investigated it and opened a pull request. The agent also used the Chrome browser and DevTools to test the feature’s operation, then moved the task to a coding channel that the engineer, product manager, and designer could follow.
During the same demonstration, a designer added a Figma file to the channel, and the agent incorporated it into the task. In the end, it published code changes alongside screenshots and a recording showing how the feature worked. Slack believes that displaying these steps in a shared place makes it easier to audit the results compared with agents that work locally on a single developer’s device.
Expanding Agents’ Presence Inside Slack
The changes are not limited to coding channels. Slack has added direct conversations with agents that work similarly to communicating with a colleague, as well as a new tab called Agents that gives each session a live status and a button to stop it. The Add to Slack mechanism also enables agents from platforms such as Lovable, n8n, OpenAI, LangChain, and Airtable to be deployed in a few steps, with OAuth setup and configuration automated.
What Changes in Practice?
Slack is betting that AI’s role in programming is shifting from improving individual productivity to supporting team workflows. According to Rob Seaman, Slack’s interim CEO, writing code is no longer the primary bottleneck; ideas, human judgment, planning, and craftsmanship have become scarcer factors. Cognition provided an internal figure indicating that the number of pull requests merged there increased tenfold in recent months, compared with an approximately 40% increase in the number of employees, while emphasizing that the figure came from the company itself.
In practice, this mechanism could allow nontechnical employees to report bugs or participate in drafting requirements, while keeping engineering review within the same process. However, this does not eliminate the need for verification; Gartner data indicates that more than 40% of agentic AI projects could be canceled by the end of 2027 because of rising costs and unclear business value. A McKinsey study also found that 62% of organizations are testing AI agents, while only about one-third have begun expanding their use, and 39% reported an impact on financial results.
Permissions and Security
According to Slack, the agent inherits the permissions of the user who invoked it and does not receive general permissions or an independent identity with broad access. What it receives from the project channel is limited to the conversation context that launched the task. Devin, meanwhile, operates inside isolated environments with limited permissions, with an option to run without an internet connection. Changes remain standard pull requests to GitHub, allowing existing release and review gates to continue.
Slack does not present collaborative channels as a complete replacement for the terminal; the company expects terminals to remain in use for deep individual work, such as redesigning a codebase’s architecture, while channels suit tasks that benefit from collective oversight, such as fixing a bug reported by the sales team. Slack plans to open the coding-channel APIs to any developer, pointing to future use cases including marketing campaigns and legal-document review.