Slack has launched Slack Code, a new feature that brings AI coding agents into dedicated project spaces where developers and other team members can plan, build, review and guide software development together. Instead of keeping coding agents inside private terminals, browser tabs or individual conversations, Slack Code creates dedicated “code channels” around specific development tasks.
The feature is part of Slack’s broader push to make AI agents active participants in workplace collaboration. Slack Code launched with integrations from Anthropic’s Claude, Cognition’s Devin, GitHub Copilot and Vercel, while OpenAI’s ChatGPT integration is listed as coming soon by Slack. The company says the same collaborative approach could eventually extend beyond software development into areas such as marketing, legal work and IT operations.
Slack Code Creates Dedicated Channels For AI Coding
Slack Code is built around the idea of giving every software project its own collaborative environment. When a user mentions a supported coding agent while discussing a development task, Slack can create a code channel dedicated to that work.
The channel brings together the AI agent, developers and other relevant team members. Instead of an agent working independently and returning a final answer, participants can follow its progress, provide additional instructions and review the output during the process.
Slack Code At A Glance
| Feature | Details |
|---|---|
| Product | Slack Code |
| Main purpose | Collaborative AI-assisted software development |
| Core workspace | Dedicated code channels |
| AI partners at launch | Claude, Devin, GitHub Copilot, Vercel |
| OpenAI integration | Coming soon |
| Code visibility | Diffs and live previews |
| Team interaction | Watch, guide, pause and review agents |
| Security | Uses existing Slack permissions and admin controls |
| Channel lifecycle | Automatically archives when work is complete |
| Initial focus | Software engineering |
The dedicated-channel model is intended to solve a problem that Slack says its own engineers encountered: standard conversation threads can become difficult to manage when AI agents perform complex, multi-step coding tasks.
Teams Can Watch AI Build Software In Real Time
One of Slack Code’s main features is visibility.
AI coding agents can produce large amounts of code quickly, but that speed can make it difficult for teammates to understand what is changing. Slack Code puts code diffs and live previews directly inside the project channel.
Team members can therefore see what an agent is building rather than waiting for a final response.
This can be particularly useful when an AI agent is working on a bug fix or user-interface change. A developer or product manager can identify a problem while the agent is still working and redirect it before the task progresses too far.
How A Slack Code Session Works
| Stage | What Happens |
|---|---|
| 1. Task identified | A team member describes a coding problem |
| 2. Agent mentioned | A supported coding agent is tagged |
| 3. Code channel created | Slack creates a dedicated project space |
| 4. Agent works | The AI plans and writes code |
| 5. Team observes | Members see progress and code changes |
| 6. Review | Developers can comment or redirect the work |
| 7. Approval | Humans review high-stakes changes |
| 8. Completion | Channel is archived while remaining searchable |
Slack says more than 70% of its code channels are created and closed within a single day, moving from an idea to a merged pull request. The company presents this as evidence that the dedicated workflow can shorten development cycles.
AI Coding Agents Become Team Participants
Slack’s approach treats coding agents less like standalone software tools and more like members of a development team.
Claude, Devin, GitHub Copilot and Vercel can be called into Slack conversations, where they can work alongside people instead of operating in isolated development environments.
This is an important shift in how workplace AI is being designed. Instead of asking an employee to leave Slack, open an AI coding tool, copy the result and then return to the team conversation, Slack wants the agent’s work to happen within the same collaboration layer.
AI Agents Available With Slack Code
| AI Agent | Company | Role |
|---|---|---|
| Claude | Anthropic | Coding and agentic development |
| Devin | Cognition | AI software engineering |
| GitHub Copilot | GitHub | Coding and development assistance |
| Vercel | Vercel | Development and web workflows |
| ChatGPT | OpenAI | Integration announced, coming soon |
The partnership model also gives Slack a way to become an interface for multiple AI providers rather than requiring businesses to commit to a single model ecosystem.
GitHub Integration Could Bring Non-Engineers Into Coding
Slack says GitHub’s integration is designed to make software development more accessible to people outside engineering teams.
A non-technical employee could describe a problem in plain language inside a code channel. An AI agent can then draft a possible fix, after which an engineer can review the proposed change.
This could change how minor software issues are handled inside organizations. Instead of every small bug or interface change immediately becoming a ticket for an engineering team, some requests could be investigated and prepared with AI before an engineer becomes involved.
Potential Workflow For Non-Technical Teams
| Traditional Process | Slack Code Approach |
|---|---|
| Employee reports issue | Employee describes issue in Slack |
| Ticket created | Agent can begin work directly |
| Engineering queue | AI investigates the problem |
| Developer starts from scratch | Agent proposes a solution |
| Developer reviews | Engineer reviews AI-generated change |
| Fix eventually ships | Approved change can move toward deployment |
The approach does not remove the need for engineers. Instead, it attempts to shift engineers toward reviewing, directing and approving AI-generated work.
Human Oversight Remains Part Of The Workflow
Slack is also building controls around AI agents because autonomous software development carries risks.
Team members can pause, redirect or stop an agent during a session. Slack says high-stakes changes can be routed to a person for approval, with human sign-off required for important actions such as merging code to production.
The platform also inherits Slack’s existing permissions and administrative controls. That means organizations do not need to create an entirely separate security and identity system for agents operating inside code channels.
Slack Code’s Oversight Model
| Control | Purpose |
|---|---|
| Shared visibility | Team can see agent activity |
| Pause | Temporarily stop an agent |
| Redirect | Change instructions during execution |
| Human approval | Review high-stakes changes |
| Existing permissions | Apply current Slack access controls |
| Audit history | Preserve completed project activity |
| Automatic archive | Prevent completed projects from cluttering channels |
This is increasingly important as AI coding agents move from generating suggestions to making larger changes across software repositories.
Code Channels Become An Audit Trail
Slack Code does not simply disappear when an AI agent completes its task.
Slack says completed code channels automatically archive while remaining searchable as an audit log. This gives teams a record of the discussion, instructions, decisions and work performed during the development process.
That could be particularly useful for businesses where developers need to understand why a particular change was made weeks or months later.
The archive also preserves context that could otherwise disappear if coding work happened in a private AI session.
Slack Wants To Reduce Context Switching
The broader objective behind Slack Code is to reduce the number of applications developers and other employees need to move between.
Slack has increasingly positioned its platform as a place where users can interact with AI agents, search company information, execute code and work with connected applications. Its July product updates, for example, positioned Slackbot as an agent capable of building dashboards, reports and other outputs directly within Slack.
Slack Code extends that strategy into software development.
Instead of Slack serving primarily as the place where employees discuss work, the company wants it to become a place where AI agents actually perform portions of that work.
Slack Code Is Part Of A Larger Agent Strategy
Slack is also introducing an Agents & Tools tab to give users a central location for managing agent conversations across DMs, threads and code channels.
That indicates Slack sees agents as becoming a persistent part of workplace communication rather than occasional assistants.
The company is beginning with engineering because coding agents are already capable of performing relatively complex tasks. But Slack says the same model could eventually be applied to other business functions.
Potential Expansion Beyond Software
| Department | Possible Agent Workflow |
|---|---|
| Engineering | Code development and debugging |
| Marketing | Campaign planning and content creation |
| Legal | Contract review |
| IT | Employee onboarding and technical support |
| Product | Research and prototype development |
| Operations | Workflow automation |
These are Slack’s stated areas of potential expansion rather than evidence that all of these capabilities are currently available through Slack Code.
Competition In AI-Powered Coding Is Intensifying
Slack Code arrives as AI coding agents become increasingly competitive.
GitHub, Anthropic, Cognition, OpenAI and other companies are building systems designed to handle increasingly large portions of software development. Slack’s strategy is different: rather than primarily competing as a coding-agent provider, it is attempting to become the collaboration layer where multiple agents can work.
That could give Slack a strategic position if companies begin using several AI coding agents simultaneously.
A development team might use Claude for one task, Devin for another and GitHub Copilot for repository-related workflows, while Slack provides the common interface for communication and oversight.
The Bigger Picture
Slack Code reflects a broader change in workplace software as AI agents move from assistants that answer questions to systems that can perform multi-step tasks. By giving agents dedicated project channels, Slack is attempting to make AI development work visible to entire teams rather than keeping it inside private coding environments.
The strategy could also strengthen Slack’s role in the enterprise AI ecosystem. By supporting agents from multiple companies, Slack is positioning itself as a neutral collaboration layer between employees and AI systems. The challenge will be maintaining security, human oversight and useful workflows as agents become capable of making increasingly consequential changes.
Looking Ahead
The success of Slack Code will depend on whether teams find shared AI development more productive than existing combinations of coding agents, terminals and collaboration tools. The ability to watch code being built, provide feedback in real time and preserve the entire process in an auditable channel could make agentic development easier for organizations to adopt.
The longer-term opportunity is much larger than coding. Slack is effectively testing a model in which AI agents become participants in everyday team workflows, with humans setting direction and reviewing important outcomes. If that approach works for software engineering, Slack could extend the same model into marketing, legal, IT and other departments, potentially turning its collaboration platform into a central operating layer for AI-powered work.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
