Salesforce brings AI coding into Slack with Slack Code, pitching “multiplayer” development for teams

Salesforce is introducing Slack Code, a new way to build software inside Slack that aims to keep AI-assisted coding work in the open, where teams can review it in real time instead of passing snippets back and forth later. The company’s pitch is that AI coding is no longer a solo activity between a developer and an agent, but a team workflow that needs the same shared context and collaboration tools teams already rely on in Slack.

In the announcement, Salesforce describes code channels as a dedicated workspace for software development collaboration with AI agents. When a user tags a coding agent, Salesforce says it can spin up a code channel for the task, allowing other participants to see the conversation, review code diffs, check live previews, provide feedback, and approve changes before they ship, all without leaving Slack.

A key element of the product concept is the company’s focus on what it calls “context.” Salesforce argues that work done between an individual and an AI agent outside Slack is difficult for others to understand or contribute to, because reviews happen later and handoffs slow down. Slack Code is designed to keep the task’s working context visible across people and agents, so the next participant can build on what’s already been captured in the channel.

Salesforce also points to practical workflow mechanics. In the example it provides, a product manager notices a bug in a channel, then asks a coding agent to help architect a fix. The agent is said to use the existing conversational context, screenshots, and documents shared with the group to propose code. An engineer can then review the diff, run a preview, and authorize the work to proceed to production through a pull request and merge flow that is initiated from within the channel.

The company emphasizes that Slack Code is meant to broaden who can participate in development work. Because the workflow happens in a tool many non-engineers already use, Salesforce says people without technical skills can help shape the output without opening a terminal. It also claims that participants can pause, redirect, or stop the agent if the work drifts, and that once the work is done, the code channel archives itself so the sidebar stays manageable while the channel history remains searchable as institutional context.

Under the hood of the user experience, Salesforce says the agent experience is being aligned with Slack’s existing security and admin controls. In code channels, it claims agents inherit Slack’s permissions and administration model from day one, without additional infrastructure or new identities for IT teams to manage. For high-stakes actions like pushing code to production, Salesforce says the agent packages its work for an expert to sign off in the channel, aiming to keep people in control while still speeding up the process.

Slack Code is also positioned as an ecosystem platform for partner agents. Salesforce says Anthropic, Cognition, GitHub, ChatGPT, Vercel, and other partners are developing for Slack Code using code channel APIs, bringing their coding agents into the same shared workspace. Looking ahead, Salesforce says the coding channel APIs will be opened to the broader developer community, so custom agents can join code channels for a wider range of work beyond software engineering.

The announcement also includes a smaller but specific user interface update related to accessibility. Salesforce says the “Buy now” button label in some circumstances can become nearly unreadable in dark mode due to contrast issues on certain devices. The company attributes the problem to a color value intended for dark-mode surfaces being applied to the button text as well, resulting in contrast well below accepted accessibility thresholds, while the button background color is not affected by theme switching.

For company context, Salesforce frames Slack Code as a step in a broader shift in how teams build software, from solo terminal work to single-player AI copilots, to what it calls multiplayer coding. The company’s central claim is that Slack is already where teams coordinate, so bringing AI agents into dedicated code channels could make it easier for knowledge to compound inside the tools where teams already operate.

Still, the company’s announcement leaves several operational questions open. It does not provide details in the release text about timelines, pricing, regional availability, supported agent types, or how code diffs and previews are rendered across different development environments. It also does not spell out how governance works for different organizational roles beyond the general statement about admin controls and expert sign-off for production moves. Customers are told in the release that pricing, packaging, and availability may vary by region and depend on customer agreements.

What to watch next is whether Slack Code’s partner ecosystem expands quickly and how broadly teams can adopt it for real development workflows, including review and approval steps that map cleanly to production release processes. Salesforce will also likely face scrutiny on accessibility and usability as more users move from individual agent chats into shared code channels, especially if agent actions can affect production systems. The rollout’s early user feedback, along with any additional documentation on supported integrations, will be an important announcement of whether “multiplayer” AI coding becomes practical at scale.