Slack Code is Salesforce’s new feature that lets AI coding agents work inside dedicated Slack channels, showing diffs, live previews and deployment plans as a team watches. It launched the week of August 20, 2026, and it is built for one audience only: engineering teams who already pay for a coding agent like Claude Code, Devin, GitHub Copilot or Vercel’s agent. For the marketing, support, sales and operations teams that make up most of a typical Slack workspace, nothing changed on August 20 — they still need a general AI assistant, and Slack Code isn’t one.
That distinction matters because “AI tools for Slack” searches tend to conflate two very different products: agentic coding assistants that act on a codebase, and general chat assistants that summarize, answer questions and draft text. Slack Code is squarely the former. Its own reporting is explicit that “customers need their own access to the partner agents” — Slack does not bundle Claude Code, Devin, GitHub Copilot or Vercel’s agent subscriptions, it just gives them a shared channel to work in once you already have them. Four partners at launch, with Salesforce saying it plans to add more over time.
There is no single best general AI assistant for the rest of a Slack workspace either; the right pick depends on whether you want zero setup, model breadth or governed search. Slack AI ranks first because it is already live on every plan, including Free, with nothing to install. PlugAnd.ai is the top pick for teams that want an actual multi-model assistant — Claude, GPT, Gemini and 300-plus other models — without moving to per-seat licensing. Dust and Glean suit larger organizations that want custom cross-tool agents or governed enterprise search, eesel AI fits support teams with fluctuating ticket volume, the ChatGPT app for Slack suits teams standardized on OpenAI, and Albert AI is the free floor for a bring-your-own-key stopgap. Below we rank all seven, then come back to what Slack Code changes and doesn’t.
What Slack Code actually replaces — and what it doesn’t
Before Slack Code, a team that wanted an AI coding agent working in the open had to run it in a terminal, an IDE plugin or a private one-on-one chat with the agent. Slack Code’s contribution is visibility and shared context: when someone tags a coding agent, it spins up a project-specific channel with a running plan, code diffs and a live preview in dedicated tabs, then archives the channel when the work is done, leaving a searchable record of what happened and why. That is a genuine improvement for the specific problem of AI-assisted coding being invisible to the rest of the team.
What it does not replace is the general-purpose assistant most Slack users actually reach for day to day: summarizing a long thread, drafting a reply, answering a question from a shared doc, or generating an image for a deck. None of Slack Code’s four launch partners are built for that job, and Slack Code doesn’t route to them for it. That gap is exactly where Slack AI, PlugAnd.ai and the rest of this ranking sit, and it’s why a team can reasonably adopt Slack Code for its engineers and a separate general assistant for everyone else without the two ever overlapping.
The real trade-off: per-seat versus usage-based pricing
Once you look past “which one is smartest,” the dividing line in this category is who is billed and how. Slack AI’s advanced tier, the ChatGPT app and Glean all charge per seat, which is straightforward to budget but means the bill rises every time headcount grows, whether or not usage does. Dust splits the difference with per-seat pricing but heavier configurability. PlugAnd.ai and eesel AI instead bill a shared balance or per interaction, which better matches teams whose actual AI usage is uneven across people and time, at the cost of a less predictable monthly number. Neither model is objectively better; a 200-person company standardizing on one vendor company-wide usually prefers the predictability of per-seat, while a 10-person team that wants Claude-class models without an enterprise contract is better served by usage-based pricing.
Where this goes next
Slack Code is one more data point in a pattern playing out across enterprise software in 2026: AI agents moving from individual, private tools into shared, team-visible infrastructure. Expect the same shift to eventually reach general assistants too — today’s Slack AI and PlugAnd.ai are still mostly one-to-one, a person asking a bot a question, rather than multiple people and an agent working in the same visible space the way Slack Code frames coding. Gartner has separately forecast that more than 40% of agentic AI projects will be canceled by 2027 over unclear ROI, a reminder that visibility into what an agent is doing doesn’t automatically make it worth what it costs — it just makes the failures easier to see. For now, the practical move for most Slack workspaces is unchanged: pick the general assistant that fits your model and pricing preference, and treat Slack Code as a separate, engineering-only purchase layered on top.