Claude in Slack can now tap into your personal apps

The most annoying part of using an AI bot in a team chat is permissions. You ask it to pull context from a project doc, but the bot is locked out because IT can't just hand everyone's Google Drive keys to a public Slack channel.

Anthropic updated Claude Tag, its beta feature that puts Claude into Slack channels, adding support for personal connectors. The feature is rolling out now for Team plan customers, with Enterprise plan availability coming soon.

Private credentials stay private in a public room

Until now, Claude in a channel could only touch shared connectors hooked up by a workspace admin. But most real work lives behind personal logins, such as your calendar, assigned CRM accounts, or a draft pitch deck only you can open.

Now, if you've connected a tool to your personal Claude account, Claude can use it when answering your specific prompts inside a channel. Nobody else in the room can piggyback on your connector to snoop around your files.

Behind the scenes, everything Claude does through your personal connector logs under your user account, matching how direct messages work. Shared channel tasks stay under the channel service account, keeping the security audit trail clean.

Auto-screening before sensitive data hits the channel

Pulling personal files into a team room comes with an obvious risk: accidentally broadcasting confidential info to the whole team.

To prevent that, Anthropic added two output modes. Review mode lets you preview and approve Claude's answer before it posts to the thread. Auto mode posts directly, but Claude will actively flag sensitive data and halt the message until you sign off. Enterprise admins will also have the option to make manual review mandatory across the board.

One clear catch is that personal connectors cannot run unattended. If you need scheduled morning briefings or an around-the-clock on-call bot triaging CI failures, you still need shared channel-level connectors.

Why it matters

We think this is one of the smartest enterprise UX moves Anthropic has made lately. The biggest blocker for workplace AI agents has never been model intelligence; it has always been permissions and compliance.

For knowledge workers, this eliminates the tedious dance of chatting with an AI in private DMs and then copy-pasting the summary into a team channel. You can now collaborate out in the open without asking IT to loosen channel-wide access controls.

By tethering tool access to the individual making the request, Anthropic neatly inherits existing enterprise role-based access control. Security teams don't have to invent a new policy engine just to let workers query their own files in public.

The friction point will likely be trust in auto mode's sensitive content detection. In regulated industries, security leads will almost certainly mandate manual review for everyone, adding an extra click to every prompt. Even so, it turns channel-based AI from a static generic helper into a practical, personal assistant that can actually see your work.