> ## Documentation Index
> Fetch the complete documentation index at: https://docs.humanintelligence.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Trusted domains

> Control which domains AI clients can connect from to reach your organization's data, and review every change.

An AI assistant reaches your people data through an MCP client, and many clients can register themselves without asking you first. When someone in your organization signs in to one, the client receives that sign-in at a domain of its own choosing. Trusted domains are the list of those domains you've approved. Trusting a domain means you're willing to let whoever controls it act on your people's behalf.

A client whose domain isn't trusted is refused.

## Included by default

Claude (`claude.ai`) and ChatGPT (`chatgpt.com`) are trusted for every organization. Connections from your own machine (`localhost`) are always allowed. You can turn Claude or ChatGPT off for your organization, and back on at any time.

Anything else — an internal tool, a vendor's agent — must be added to your list.

## Adding a domain

**1. Open Trusted Domains**

On the Connectors page, open the **Trusted Domains** tab. If you don't see it, ask a Platform Admin.

**2. Click Add Domain**

Enter the domain and, optionally, a note on what it is and who asked for it. The note is shown beside the domain.

**3. Save**

The domain is trusted immediately and the change is recorded in the Audit Log.

Enter the domain only, with no `https://` and no path.

| Format | Example | Trusts |
| - | - | - |
| Single host | `apps.example.com` | That host only |
| Wildcard | `*.example.com` | Every subdomain, such as `apps.example.com`, but not `example.com` itself |
| App scheme | `cursor:` | A desktop app that signs in through its own link scheme |

Wildcards that would cover a whole top-level domain (`*.com`) or a shared hosting platform are refused. Enter the specific host instead.

> [!TIP]
>
> Prefer a single host to a wildcard. Trust the narrowest domain that covers the client.

## Seen but not trusted

This section lists clients that tried to connect in the last 30 days from an untrusted domain and were refused. A domain shown here asked one of your people for access to your data. Add only the ones you recognize.

For each domain you see the people who tried, the number of attempts, and when it was last seen. Click **Trust Domain** to open the add dialog with the domain and a note already filled in.

## Editing and removing domains

Use a row's actions menu to **Edit note** or **Remove domain**. Removing a domain disconnects everyone in your organization who connected through it. You can add it again later.

Domains under **Included by default** can't be deleted, only turned off.

## Auditing changes

The **Audit Log** tab lists every change to your trusted domains, newest first. Entries are never edited or removed.

| Column | Shows |
| - | - |
| **When** | Date and time of the change |
| **Change** | Trusted, Note edited, Stopped trusting, Turned on default, or Turned off default |
| **Domain** | The domain affected |
| **Details** | The note added, edited, or removed |
| **By** | The person who made the change |

Search by domain, person, or type of change, for example to find who stopped trusting Claude. A filtered view has its own link you can share. The log shows the most recent 500 changes; contact us for older entries.

> [!NOTE]
>
> The Audit Log records changes to the list, not individual sign-ins. To see who tried to connect from an untrusted domain, use **Seen but not trusted**.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.