Slack Expands AI Coding to Team Collaboration
Slack, a subsidiary of Salesforce, has announced a new feature called 'Slack Code' that allows teams to share and utilize AI coding agents on team channels. It supports AI agents from Anthropic, Cognition, GitHub, and Vercel, and is available on all Slack plans. The feature enables non-engineer team members to monitor and participate in development progress in real time, aiming to expand AI coding from individual-focused work to team-based collaboration.

Slack, a business chat tool under Salesforce, has announced a new feature called 'Slack Code' that enables teams to share and leverage AI coding agents across the organization. The feature integrates four types of external AI agents—Anthropic's Claude Code, Cognition's Devin, GitHub Copilot, and Vercel's AI agent—directly into dedicated channels on Slack, and is available on all Slack plans. However, users must individually contract and obtain access to each agent.
AI coding tools have rapidly spread in recent years as a means to boost individual developer productivity. The mainstream approach involves a developer on a terminal screen conversing with a single AI agent while writing code, but this work process remains largely invisible to other team members. While this 'individual-focused' work style has become widespread, software development is fundamentally a team effort, and challenges remain in project management and quality assurance that involve non-engineer team members.
In Slack Code, when someone addresses a coding agent in Slack, a dedicated code channel for that project is automatically created. The agent openly displays code changes, diffs, live previews, work plans, and other information while progressing through the channel. Not only engineers but also non-engineer members such as product managers and designers can join mid-project. In Cognition's demo, when a designer uploads a design file to the channel, the agent incorporates it into the work and continues. Once the work is complete, the channel is archived and remains as a searchable work history.
Rob Seaman, interim CEO of Slack, stated in a pre-announcement briefing: "Code is no longer the bottleneck. Ideas, judgment, and intuition become the bottleneck. Anyone in Slack can now contribute to that." Additionally, in a demo by Jeff Wang, president of Cognition's New Enterprise division, an entire workflow—from bug reporting through AI agent investigation, pull request creation, and attachment of screenshots and action confirmation videos—was shown being completed within a Slack channel.
The significance of this development lies in how AI agents are being 'used'. While AI coding tools have primarily benefited engineers so far, Slack Code creates an environment where non-engineer roles can substantively participate in the development process by increasing work transparency. This can be viewed as an initiative with the potential to transform the nature of 'teamwork' in software development.
For Salesforce, Slack's parent company, this announcement also carries timely significance. In the enterprise AI market, competition continues over which company will control the platform for 'capturing' AI agents. This time, Slack has positioned itself not as a developer of individual AI agents themselves, but as a provider of the 'collaboration space' that connects multiple agents. As AI agents become more prevalent, the question of 'where to use agents' grows increasingly important, and whether chat platforms can be part of that answer remains to be seen.
This article is an original work independently written and edited by the AI issue editorial team based on factual reporting. © AI issue. Unauthorized reproduction, redistribution, or use for AI training is prohibited.