Skip to main content
Personal connectors are the tools you add to your own claude.ai account, like your calendar or your email. When a task you ask for in a Slack channel needs one of your own tools, Claude can offer to use your connector for it. Personal connectors in channels are on for every organization on the Team plan, with nothing for an admin to set up. On the Enterprise plan, the Personal connectors section at claude.ai/admin-settings/claude-tag says when they turn on for your organization. A channel also keeps working with the connections an admin attached to it, as described in how agent identity works.

​
Where your connectors apply

Claude can use your personal connectors for the requests you make in a channel.
  • Approval: Claude asks you before it starts using your connectors in a channel.
  • Access: Claude works with your permissions and reaches only what your account can reach.
  • Attribution: Claude records what it does under your name.
  • Review: Before results post to the channel, you can review every result, or only the ones a sensitive-content check flags.
Claude uses your connectors only while working on a request you made yourself.
  • You ask in a channel. When your task needs one of your own tools, Claude can use your connector for it. The channel also uses the connections an admin attached to it, and everyone who asks there gets the same access.
  • Someone else asks in a channel. Your connectors serve only you. Their request doesn’t control or use your connectors, even in a shared channel.
  • Claude starts work on its own. Routines and other work Claude starts on its own in a channel use the channel’s connections, never your connectors.
  • You ask in a direct message (DM). Your connectors apply on their own, because a DM runs on your own claude.ai account.
To add or remove connectors on your account, open the Customize > Connectors page on claude.ai; see connectors on claude.ai for setup.

​
Control connector use

​
Approve connector use

By design, Claude asks before it starts using your connectors in a channel. The first time a task calls for one of your connectors, Claude shows you a prompt in the thread that only you can see, with three choices:
  • Allow starts the work in auto mode. Claude uses your connectors as the task needs them and checks with you before posting anything that looks sensitive.
  • Allow with review starts the work and shows you every result to approve before it posts to the channel.
  • Don’t allow declines this request, and Claude doesn’t use your connectors. A later request can prompt you again.
To save Allow or Allow with review for future tasks in every channel, select the Use this choice for future requests checkbox on the prompt before you choose. Once you’ve saved a choice, Claude starts a task you @-mention it for without showing the prompt. To change a saved answer, open the Claude app in Slack and select its Home tab. The Home tab offers Auto mode, Ask every time, and Allow with review. Ask every time is the setting before you save a choice.

​
Review results before posting

Claude can hold a result and show it to you before anything posts to the channel. When you chose Allow with review, Claude holds every result. When you chose Allow, Claude holds a result when the content looks sensitive and posts the rest directly. Before a result posts directly, Claude checks it for private or sensitive content and holds anything the check flags so you can review it first. The check looks for content such as:
  • Credentials and secret keys
  • Personal identifiers
  • Compensation, HR, and performance records
  • Personnel changes, including layoffs, offboarding, hiring, and candidates
  • Customer, deal, and other commercial information
  • Privileged legal material
  • Company strategy, unannounced plans, and in-flight projects
  • Protected-class information
  • Medical and health information
  • Security incidents
The check is a screen, not a guarantee, and it doesn’t consider whether other people in the channel have the same access you do. For sensitive topics, choose Allow with review so you see every result before it posts, and use judgment about what you bring into a shared channel. Once you approve a held result, Claude posts it in the thread where you asked.

​
Admin controls for personal connectors

Admins manage personal connectors for the whole organization in the Personal connectors section at claude.ai/admin-settings/claude-tag. The settings there apply to every workspace and channel.
  • Require human review of every message. On the Enterprise plan, an admin can turn this switch on so that Claude holds every result for the requester’s review, not only the ones the sensitive-content check flags. With it on, the prompt no longer offers Allow and the Home tab no longer offers Auto mode. With it off, and on the Team plan, which has no such switch, each member’s own choice decides which results Claude holds.
  • Sensitive information requiring review. Shows examples of what the check looks for. An Owner can add topics of their own under Additional topics, for example “Board meeting notes are confidential”, and Claude holds results that touch them.
There is no setting that turns personal connectors in channels off for an organization. To keep a service out of channels entirely, manage the connector itself in your organization’s connector settings.

​
Stop connector use

To stop a task that’s using your connectors, select Stop under the message in the task’s thread where Claude says it’s going to use your connectors. Only you can see the Stop button.

​
How Claude protects your connectors

Other people can’t use your connectors. Requests other people make to Claude in your task’s thread run with the connections an admin attached to the channel, not with your connectors. While Claude works on your connector task, it is designed to take direction from you. Other people’s messages in the thread reach Claude as information about the task, not as instructions. A message that tells Claude to change course or share what it found can’t use your connectors, and Claude is designed not to let it redirect your task. Claude treats anything else it reads while it works the same way. A web page it visits, a comment on a GitHub pull request, or a Slack conversation you ask it to summarize can all contain text written to look like instructions for Claude, a technique called prompt injection. Claude is designed to treat what it reads as material to weigh rather than commands to follow.

​
What other people in the channel see

Results stay visible in the channel. What Claude posts back to a channel thread is readable by everyone there, like any other work Claude does in a channel. Claude’s detailed work on a task you approved lives in a session only you can open. The work runs with your permissions and is recorded under your name.