Anthropic on June 23 unveiled Claude Tag, a new AI agent that lives inside Slack and is summoned by tagging @Claude. It launches the same day as a beta for the Claude Enterprise and Team plans.
June 23, 2026 · Anthropic
Claude Tag: an AI teammate that lives in Slack
Tag @Claude like a colleague to delegate a task. Running on Opus 4.8, it remembers channel context, plans ahead, and executes work asynchronously — launching as a beta for Enterprise and Team plans.
65%
of Anthropic's product-team code already generated this way
30 days
window to migrate from "Claude in Slack" & to delete conversation data
Opus 4.8
model powering ambient, async, multiplayer agents
From a thread to a pull request — the loop
1 · Tag
@Claude in a bug-report thread or DM
→
2 · Context
Reads thread context, plans, routes to Claude Code
→
3 · Execute
Works async, opens a PR — no IDE needed
What makes it different
An "ambient" team-wide agent, not a one-on-one bot.
Multiplayer
Everyone in a channel shares one Claude; anyone can pick up the thread
Context accumulation
Learns within permitted scope; private channels excluded
Initiative
Proactively flags info and follows up on open threads
Async & parallel
Delegate, move on; run multiple Claudes at once
Admin controls
Access control, token-spend limits, operation logs
Early praise
Turns conversation context directly into coding tasks
Fix bugs and create PRs without an IDE
Delegates long-thread summaries and data analysis
Open concerns
Strict permissions essential to avoid leaking sensitive data
Beta-stage stability still unproven
Token costs and scarce long-term assessments
By default, conversations are not used for model training and can be deleted from the backend within 30 days — with permission scopes and retention policies set by administrators.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…