Anthropic Launches Claude Tag — an Always-On AI Teammate That Lives in Slack
Anthropic introduced Claude Tag, a new way to put Claude directly into Slack as a team member. Add it to selected channels, connect tools, and anyone can @Claude to delegate work. It builds context from the channels it sits in, plans tasks to run later, and works asynchronously — you can hand it a project and it’ll grind on it over hours or days, even scheduling follow-up work for itself.
Everyone in a channel shares one Claude identity, so a teammate can pick up where a colleague left off and see exactly what Claude has been doing. An optional “ambient” mode lets it proactively flag things it thinks you should know. It’s in beta for Enterprise and Team customers, with access to sensitive data and tools tightly scoped. Anthropic says 65% of its own product team’s code is now written by an internal version of the tool.
Claude Design Settles Down for Everyday, On-Brand Work
Anthropic shipped an update to Claude Design aimed at daily use: it now stays on brand by building with your own design system components, and moves more cleanly between the visual canvas and code. The pitch is consistency — designs that look like your product, not a generic template — and a steadier hand-off from concept to implementation.
It’s a quiet but telling release. As Claude pushes deeper into enterprise workflows, “respects your design system” is exactly the kind of unglamorous reliability that decides whether a tool gets adopted by real teams or stays a demo.
Centrally Managed Authorization Lands for MCP Connectors
Anthropic added enterprise-managed authorization for MCP connectors, starting with Okta. Admins can provision a connector once and users get zero-touch access on first login — no more chasing individual OAuth grants. The beta spans Claude chat, Claude Code, and Cowork for Team and Enterprise plans.
Pair it with the newly GA Workload Identity Federation on the Claude Platform, which lets organizations authenticate workloads without static API keys, and the direction is clear: Anthropic is building the identity and access plumbing enterprises require before they wire Claude into production systems at scale.
Managed Agents Get Schedules and Vaulted Environment Variables
Claude Managed Agents can now run on a schedule and store environment variables in vaults to authenticate CLIs and other tools. Together those turn one-off agent runs into durable, recurring automation — a nightly report, a scheduled data sync, a periodic cleanup — without leaving credentials sitting in plaintext.
It builds on the recent move that lets Managed Agents execute inside a sandbox you control and reach your private MCP servers, keeping both the execution environment and the services it touches within enterprise boundaries.
Claude Takes Two Separate Outages in One Day
June 23 was a rough day for reliability. Anthropic opened a first incident at 06:28 UTC tied specifically to Opus 4.8, resolving it after more than two hours at 08:45 UTC. Then a second, broader incident hit from roughly 14:08 to 15:33 UTC, with elevated error rates across multiple models returning to normal by 16:44 UTC.
The errors touched claude.ai, the API, Claude Code, the Console, and Cowork — but notably not Claude for Government, which stayed online through both. Two multi-hour incidents in a single day, the morning one pinned to the flagship model, will sharpen questions about capacity headroom even after recent compute deals.
Fable 5 Still Suspended on Day 12 as ID Verification Nears
Community trackers continue to read “not back” for Fable 5 and Mythos 5, twelve days after the US export-control directive took them offline. ID verification — not a formal lifting of the order — remains the most-watched off-ramp: Anthropic’s privacy policy now enumerates government ID and biometric data, with verification going live July 8.
Analysts point to two dates that will likely decide the shape of any return: the July 8 verification rollout and an August 1 deadline tied to the covered frontier-model framework. A US-citizens-only restoration that satisfies the directive without a government-level agreement is the scenario most observers now expect.
Claude Tag Is the Quiet Pivot From “Tool” to “Coworker”
Strip away the Slack packaging and Claude Tag is a bet about where AI value actually lands: not in a chat window you visit, but in a persistent presence that sits in your channels, holds shared context, and does work while you sleep. A single identity per channel, asynchronous task-running, self-scheduled follow-ups — this is Anthropic designing Claude to behave like a headcount, not a feature.
The “65% of our product team’s code” stat is the proof point and the provocation. It signals real internal leverage, and it reframes the buyer’s question from “which model is best” to “how many roles can this absorb.” For a company in IPO diligence, an always-on enterprise teammate — metered, scoped, and sticky — is a far better growth story than another benchmark win. The outages are the counterweight: you can’t sell a teammate that calls in sick twice in one day.