Anthropic has built a Slack integration that turns Claude into an always-on agent watching your channels and ready to jump in. Instead of a simple chatbot you summon, this setup runs quietly in the background, scanning for keywords, project updates or support requests. When it spots something it thinks you’ll want to act on—like a stalled thread or repeated question—it pops up suggested replies, draft messages or channel summaries. You can tweak its sensitivity, pick which channels it monitors and set rules for when it should intervene. Under the hood, Claude’s running with a rolling context window that captures recent messages across multiple threads. It can pull in attachments, links and emoji reactions to build a fuller picture of what’s going on. You can ask it to generate quick status updates, extract action items from a meeting discussion or spin up a draft announcement. Early testers say it cuts down on manual summarizing and ticket-triaging work, though they’re still debating how much control to give the agent versus relying on human judgment. Privacy and security got top billing in the announcement. All data stays within your Slack workspace unless you explicitly send content back to Anthropic’s servers for fine-tuning. Administrators can turn channels on or off, set guardrails around sensitive terms and review every suggestion before it posts. Anthropic is positioning this as a preview for enterprise customers right now, with broader rollout slated for later in the year. It’s a direct play against Microsoft’s Teams Copilot and Google’s Duet AI in the race to bake smart assistants into popular collaboration tools.