Slack is on a mission to become an agentic platform. On August 20, the company launched Slack Code, an AI coding environment that will enable teams to use coding agents like Claude Code and GitHub Copilot directly in Slack.
According to its announcement blog post, Slack Code will launch with dedicated, project-based code channels where teams can manage AI agents and share context in real time. Users can also tag coding agents to spin up code channels from any conversation.
The release comes as part of a broader effort by the platform to embrace AI agents. Just recently, in March, the company rolled out a revamped version of Slackbot, positioning it as a personal AI agent that takes context from messages, files and channels. Together, Slack Code and Slackbot provide employees and developers with access to agentic capabilities in a single interface.
The AI Code Race
Coding agents have become integral to modern software development. According to an Anthropic survey of 500 technical leaders in the U.S., nearly 90% of organizations use AI to assist with coding today. Tools like Claude Code, GitHub Copilot and Codex have all accelerated the velocity of code development, and Slack is aiming to position itself as a key channel for interacting with these tools.
Katie Steigman, VP of product at Slack, told me in an interview that Slack started building Code about three months ago to try and better integrate agents into the platform. Initially, development began with just two engineers. “It was a very scrappy, very iterative effort.”
For Steigman, the approach goes beyond vibe coding. “I would say it’s the next evolution of vibe coding because it’s not just me being like, ‘let me make some slop,” Steigman said. “Now I can actually tag in a designer to be like, ‘hey, like can you share the components you think the agent should use for this?’ And I can tag in engineers, and the engineers can slap my hand and be like, ‘I will work with the agent to clean this up’, and then we can ship something that’s like more viable than it was before.”
Steigman notes that Slack is positioning itself as a “neutral platform” for agentic work, offering an experience that is designed to be independent of the model or agents chosen by the end user. She highlights that people want the flexibility to switch models and model providers.
For instance, users can tag in coding agents like Claude or Cognitions Devin, and then the tagged agent will spin up a code channel to perform the assigned task without leaving Slack. She added that customers including Anthropic, OpenAI and Cognition are “doing most of their coding in Slack,” clarifying that all initial launch partners were implementing code channels internally and providing feedback.
Agentic Workspaces
With Slack facing steep competition from Microsoft Teams to be the communication platform of the enterprise, greater integration with frontier AI agents has the potential to help capitalize on the demand for AI coding tools by offering an agentic workspace.
“Everyone is getting into the coding game, and companies with that capability are becoming valuable, as the SpaceX and Cursor deal has shown. Slack’s Code will need developers to move away from their current platform of choice and adopt new workflows. That is a question that is difficult to answer and will need a lot of work with persuasion,” Shashi Bellamkonda, principal research director at Info-Tech Research Group, told me via email.
Bellamkonda says that it is a significant benefit in granting agents access to information without employees needing to transition between multiple applications, but also notes challenges, such as to whether agents get correct and up-to-date information. Likewise, increasing the speed and volume of AI production might put organizations on the verge of creating more technical debt and vulnerabilities.
Governance has the potential to be a challenge to if agents act in unaligned ways, though users have the option to stop agents mid-task, alongside Slack’s existing permissions and admin controls. Similarly, before pushing code to production, agent packages are held for an expert to sign off in the channel in an attempt to keep humans in the loop.
It’s worth noting that Slack also plans to expand the solution, sharing in the announcement that soon code channel APIs will be open to the broader developer community, who will be able to bring any custom agent into a code channel.
A New Era Of Coworking
Enterprise AI adoption has largely been siloed across third-party providers. “The companies that are reaping the biggest returns from their AI investments are doing so when they can bring AI into the organization as a team and not have it be a solo-player sport,” Ryan Gavin, CMO of Slack, told me in a video interview before the launch of Code. “All these agents need a home and a place to work with the teammates, and that’s Slack.”
At its core, Slack is aiming to become the operating layer for enterprise agents, with Slackbot providing enterprise search capabilities and Code providing software development functionality. The organization told me that 77% of Fortune 100 companies use Slack, while its website claims the tool is used by over 750,000 organizations.
The approach does have some limitations. Back in March, Gartner released a report suggesting that advancements to Slackbot pivoted Slack into an AI work hub for collaboration, but argued that it depends heavily on Agentforce, making it less competitive compared to other popular enterprise AI assistants.
In any case, in a market where so many AI model vendors are vying to be the definitive enterprise solution, Slack is instead positioning itself as a key enabler for accessing third-party AI products within a single workspace. It’s an approach that will serve as a key differentiator between Slack, Teams and the underlying Office 365 ecosystem.