Slack Code brings AI coding agents into team channels
AI coding has been a solo sport. One developer, one agent, one terminal window, invisible to everyone else on the team. Slack Code is Salesforce’s attempt to change that.
What Launched
Slack announced Slack Code, a new product that embeds AI coding agents directly into dedicated Slack channels. Supported agents at launch include Anthropic’s Claude Code, Cognition’s Devin, GitHub Copilot, and Vercel’s agent. It’s available on any Slack plan, though customers need their own access to each partner agent.
When someone tags a coding agent from any conversation, the agent spins up a project-specific code channel. It does the work in the open with code diffs, live previews, and a running plan visible in dedicated tabs, then archives the channel when done, leaving a searchable audit trail.
How It Works in Practice
Cognition’s Jeff Wang, president of new enterprise, demoed a live workflow: a broken feature gets reported in an engineering channel, Devin picks it up, investigates using Chrome and DevTools, and opens a pull request, tagging the relevant code owner automatically. A designer can drop a Figma file mid-task and the agent incorporates it without losing context. The agent finishes by posting code changes alongside screenshots and a recorded demo.
Beyond code channels, Slack is also shipping agent DMs, a new Agents tab with live status and a stop button, and an “Add to Slack” flow supporting Lovable, n8n, OpenAI, LangChain, and Airtable with OAuth and configuration automated.
The Numbers Behind the Pitch
Cognition reported its internal merged PR count went up 10x over recent months while headcount grew roughly 40 percent. Slack’s interim CEO Rob Seaman framed the strategic argument directly: writing code is now the cheap step, and human judgment is the scarce one.
The skeptics have data too. Gartner predicted more than 40 percent of agentic AI projects will be canceled by end of 2027. McKinsey’s most recent State of AI survey found 62 percent of organizations are experimenting with AI agents, but only about a third have started scaling, and just 39 percent report any bottom-line impact.
Permissions Model
Agents inherit the permissions of the human who invokes them and nothing more. Seaman described it as: “Everything is done on behalf of the user, using the user’s ACLs, both in Slack and in the systems that they’re connecting to. There’s no god permissions or bot-level permissions.” Devin runs in isolated sandboxes with minimum viable access, including an optional mode with no internet access. Because agents produce standard pull requests into GitHub, existing review gates still apply.
The Operator Angle
If you run a small team where non-engineers report bugs or spec features, this is the pitch worth watching. The plan is to open code channel APIs to any developer, with marketing campaigns and legal document reviews named as next use cases after software engineering.
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
