--- title: Slack turns coding agents into channel-dwelling teammates description: Slack has launched Code channels, dedicated spaces where humans and AI coding agents like Claude, Copilot, and Devin plan, write, review, and ship code... canonical: https://daily.dev/posts/slack-turns-coding-agents-into-channel-dwelling-teammates-2qzjj400w twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Slack turns coding agents into channel-dwelling teammates | daily.dev og:description: Slack has launched Code channels, dedicated spaces where humans and AI coding agents like Claude, Copilot, and Devin plan, write, review, and ship code... og:url: https://daily.dev/posts/slack-turns-coding-agents-into-channel-dwelling-teammates-2qzjj400w og:image: https://api.daily.dev/og/posts/2QzJj400W.png og:image:alt: Slack turns coding agents into channel-dwelling teammates og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Slack turns coding agents into channel-dwelling teammates **[Trends](https://daily.dev/sources/trends)** · 3 min read · 0 upvotes · 0 comments ## Summary Slack has launched Code channels, dedicated spaces where humans and AI coding agents like Claude, Copilot, and Devin plan, write, review, and ship code together. Tagging an agent spins up a project channel, and agents appear in DMs and the Agents tab like a coworker, though human sign-off remains required for key steps. Reaction has been largely positive, framed less as a flashy AI feature and more as sensible workplace design that gives non-engineers visibility into code as it happens, while some skepticism centers on how loosely 'key steps' requiring sign-off might get defined under deadline pressure. ## Content Slack just shipped Slack Code, and the pitch is simple: tag an agent, get a dedicated channel with a plan, live diffs, PR details, and an HTML preview that updates in real time. Claude, Devin, GitHub Copilot, Vercel's agent, and eventually ChatGPT can all spin these up. Nobody but agents can create one. Humans can steer, pause, or kill the agent mid-task, and once the work's done the channel archives itself into a searchable audit log. It's free on any Slack plan, though the agent vendors bill separately for usage. The reception is split between genuine enthusiasm and a raised eyebrow at the timing. On the enthusiast side, developer Ben Holmes calls it "an unpopular take" that Slack is actually a decent place to work with agents, since you can watch how better engineers write their prompts, pull designers in before shipping something ugly, and review screenshots the agent sends without leaving the thread. Box CEO Aaron Levie frames it as a form-factor problem finally getting solved: agents doing longer tasks with less oversight can now surface inside tools people already live in, rather than needing constant babysitting in a separate window. Then there's Gergely Orosz, whose reaction is basically relief mixed with mild shade: "FINALLY, Salesforce/Slack noticed... Better late than never!" That captures the general vibe among developers who've watched Anthropic's Claude Tag and Block's Buzz already stake out this territory. Slack showing up now reads less like innovation and more like catching up to a race that started without it. Reportedly there was internal friction too: Anthropic's presence in the launch lineup sat awkwardly next to Salesforce's own Agentforce ambitions. The skepticism isn't just about timing. Analysts covering the launch point to real friction risks: Slack's interruption-heavy culture doesn't obviously mesh with focused coding work, developers may just hide messy prompt iterations and paste clean results, channels could sprawl unmanageably at scale, and channel membership doubling as repo access is a privilege-escalation headache waiting to happen. The likely outcome, per that read, is a hybrid: shared channels for small cross-functional asks, private agent sessions for anything resembling a real refactor. Worth noting the guardrails exist because they've been needed before. Slack explicitly built in the ability for any channel member to pause or kill an agent mid-task, a direct response to past incidents of agents wiping codebases. Merges to production still require a human to sign off. Slack's also opening the code-channel APIs so agents from other domains, legal, marketing, whatever, can join the same pattern. Whether that ends up feeling like "multiplayer coding" or just Slack notification fatigue with extra steps is still very much up for debate. ## Questions this post answers ### What is Slack Code and how does it let AI coding agents work inside Slack? Slack Code is a Slack feature where tagging a coding agent spins up a dedicated channel containing a plan, live diffs, PR details, and a real-time HTML preview. Only agents can create these channels, humans can pause or kill an agent mid-task, and completed channels archive into a searchable audit log. Supported agents include Claude, Devin, GitHub Copilot, and Vercel's agent, with ChatGPT support planned; it's free on any Slack plan, though agent vendors bill usage separately. _Developers weighing where to run AI coding agents can compare approaches like this via daily.dev._ ### What safety controls does Slack Code have to prevent an AI agent from damaging a codebase? Any member of a Slack Code channel can pause or kill an agent mid-task, a guardrail added specifically in response to past incidents where agents wiped codebases. Merges to production still require explicit human sign-off, meaning the agent can propose and iterate but cannot push changes live without a person approving them. _Teams adopting coding agents can track safety practices like these through daily.dev._ ### What are the main risks or downsides people see with running AI coding agents in Slack channels? Critics point to several friction risks: Slack's interruption-heavy culture doesn't suit focused coding work, developers may hide messy prompt iterations and only paste clean results, agent channels could sprawl unmanageably at scale, and channel membership doubling as repo access creates a privilege-escalation risk. The likely resolution is a hybrid model, using shared channels for small cross-functional tasks and private agent sessions for larger refactors. _Anyone evaluating agent-in-chat workflows can weigh these trade-offs alongside similar coverage on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 2 discussions and 8 comments across x (as of 2026-09-13). **TL;DR:** Reactions are mostly enthusiastic about agents working inside shared channels, though a few raise concerns about trust, permissions, and verifying what agents actually did. **Sentiment:** 40% positive · 45% mixed · 15% skeptical **The case for** - Some see agents collaborating heavily inside shared channels as impressive and a glimpse of the future of in-line agent workflows. - There's a view that seamless, low-friction agent UX (working silently inside existing tools) drives natural adoption. **The pushback** - One commenter notes that stated permissions (e.g. 'read-only') can be undermined when an underlying MCP server actually writes using the user's token, so the UI must reflect what actually happened, not just configured intent. - Trust is flagged as a challenge — verifying agents did the job right without manually checking every output. **By community** - x (mixed): Replies mix excitement about agents collaborating in shared channels with pointed concerns about permission mismatches and trust in unverified agent output. **Hottest debate:** Whether the interface can be trusted to reflect what an agent actually did versus what permissions it was merely granted. **Open questions** - How should interfaces reliably surface enforced actions versus configured permissions when agents use underlying tools like MCP servers? - How can trust be established that an agent completed a task correctly without manually reviewing every output? **Highlights** > @Secondmindsys @levie Mine said read-only on the repo. The MCP server it called ran under my token and wrote. That pane has to render what the enforcing process reports back, not the config the UI received at boot. > — [killix on x](https://x.com/killix/status/2090659133309464721) > @levie The best agent UX might be no UX at all. When agents handle the work silently inside tools people already use, adoption happens naturally. The challenge is building trust that the agent did the job right without needing to check every output. > — [belajar\_ai\_noob on x](https://x.com/belajar_ai_noob/status/2090657246388932710) > @levie As agents work for longer stretches, the interface starts becoming shared work state rather than just conversation. The surrounding software has to make visible what the agent is doing, what it’s allowed to do, what changed, what needs review, and what evidence supports > — [Secondmindsys on x · 1 comments](https://x.com/Secondmindsys/status/2090655597037343201) > @levie it is amazing to watch @getskydive agents collaborate within our @footworkvc slack > — [nbt on x · 1 comments](https://x.com/nbt/status/2090656214803964220) **Source threads** - [x](https://x.com/GenAISpotlight/status/2090427134393163999) · 0 points · 0 comments - [x](https://x.com/levie/status/2090655109994979528) · 0 points · 8 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#code-review](https://daily.dev/tags/code-review), [#vibe-coding](https://daily.dev/tags/vibe-coding), [#slack](https://daily.dev/tags/slack) [View this post on daily.dev](https://daily.dev/posts/slack-turns-coding-agents-into-channel-dwelling-teammates-2qzjj400w) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Slack turns coding agents into channel-dwelling teammates","url":"https://daily.dev/posts/slack-turns-coding-agents-into-channel-dwelling-teammates-2qzjj400w","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/slack-turns-coding-agents-into-channel-dwelling-teammates-2qzjj400w"},"datePublished":"2026-08-20T15:08:45.267Z","dateModified":"2026-09-13T19:28:06.440Z","description":"Slack has launched Code channels, dedicated spaces where humans and AI coding agents like Claude, Copilot, and Devin plan, write, review, and ship code...","isAccessibleForFree":true,"articleSection":"Trends","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/slack-turns-coding-agents-into-channel-dwelling-teammates-2qzjj400w","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,code-review,vibe-coding,slack","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"Slack turns coding agents into channel-dwelling teammates"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/slack-turns-coding-agents-into-channel-dwelling-teammates-2qzjj400w#faq","mainEntity":[{"@type":"Question","name":"What is Slack Code and how does it let AI coding agents work inside Slack?","acceptedAnswer":{"@type":"Answer","text":"Slack Code is a Slack feature where tagging a coding agent spins up a dedicated channel containing a plan, live diffs, PR details, and a real-time HTML preview. Only agents can create these channels, humans can pause or kill an agent mid-task, and completed channels archive into a searchable audit log. Supported agents include Claude, Devin, GitHub Copilot, and Vercel's agent, with ChatGPT support planned; it's free on any Slack plan, though agent vendors bill usage separately. Developers weighing where to run AI coding agents can compare approaches like this via daily.dev."}},{"@type":"Question","name":"What safety controls does Slack Code have to prevent an AI agent from damaging a codebase?","acceptedAnswer":{"@type":"Answer","text":"Any member of a Slack Code channel can pause or kill an agent mid-task, a guardrail added specifically in response to past incidents where agents wiped codebases. Merges to production still require explicit human sign-off, meaning the agent can propose and iterate but cannot push changes live without a person approving them. Teams adopting coding agents can track safety practices like these through daily.dev."}},{"@type":"Question","name":"What are the main risks or downsides people see with running AI coding agents in Slack channels?","acceptedAnswer":{"@type":"Answer","text":"Critics point to several friction risks: Slack's interruption-heavy culture doesn't suit focused coding work, developers may hide messy prompt iterations and only paste clean results, agent channels could sprawl unmanageably at scale, and channel membership doubling as repo access creates a privilege-escalation risk. The likely resolution is a hybrid model, using shared channels for small cross-functional tasks and private agent sessions for larger refactors. Anyone evaluating agent-in-chat workflows can weigh these trade-offs alongside similar coverage on daily.dev."}}]} ```