> ## Documentation Index > Fetch the complete documentation index at: https://docs.omni.co/llms.txt > Use this file to discover all available pages before exploring further. # AI connectors > Use AI connectors (MCP servers) to give Omni AI features access to your other apps. AI connectors let the [Omni Agent](https://docs.omni.co/ai/chat.md) reach beyond Omni and into the other apps your team works in, so it can pull in context from those apps while you chat. Any app with an MCP server can be added as a connector, and after it's added, its tools are available to the agent in your chats. ## How it works A connector links Omni to another app's MCP server. When you add a connector, its tools become available to Omni's AI features, such as the [Omni Agent](https://docs.omni.co/ai/chat.md). ### Connection methods When you add a connector, you choose how people in your organization access the server: | Method | How it works | Available in embed? | | - | - | - | | **No account needed** | The server is open to everyone in the organization. No sign-in is needed. | **No** | | **Each person connects their own account** | [Each person signs in to the app](#connect-your-account). The connector's tools act as that person. | **No** | | **One shared API key for everyone** | All tool calls use the API key that you add. | **Yes** - [read this first](#embed) | With the exception of **No account needed**, the connection method can't be changed after you add a connector. Remove the connector and add it again to change its connection method. ### Connectors in embedded sessions > [!WARNING] > > Every connector that uses a shared API key is available to all users in your [embedded Omni instances](https://docs.omni.co/embed.md). You can't limit a shared-key connector to internal users only. A shared-key connector makes every tool call with your organization's API key, not with the embed user's own account. This means embed users can access anything in the connected app that the key can access. This is especially important if your organization uses Omni for both internal BI and embedded analytics. For example, if you add your internal wiki as a shared-key connector for your team, your embed users can also use it. Omni plans to add functionality that allows you to choose if a connector is available in embedded sessions. Until then, to limit what embed users can access: * **Restrict the API key's access.** Scope the API key to only have access to data that embed users are allowed to see. * **For an app that embed users must not access, don't use a shared API key.** Connectors that use other connection methods aren't available in embed. ### Session-level control In the Omni Agent, each user can [turn a connector's tools on or off](https://docs.omni.co/ai/chat.md#connectors) for their own sessions. This doesn't affect other users. ## Requirements **Organization Admin** permissions are required to manage custom AI connectors. ## Adding connectors The steps for adding a connector differ depending on how you want users to connect: > [accordion] No account needed > > Open to anyone in the organization > > 1. Navigate to **AI Hub > Connectors**. > 2. Click **Add connector**. > 3. Fill in the following: > * **Name** - Enter a name for the connector. This is what displays to users throughout the Omni UI. > * **Server URL** - Enter the server URL. This can't be changed after the connector is created. > * **How do people connect?** - Select **No account needed**. If the connector turns out to need sign-in, Omni switches this setting to per-person for you. > 4. **Optional**: Set **Advanced** options: > * **Transport** - Streamable HTTP (most servers) or SSE (older servers). Streamable HTTP is the default. > 5. When finished, click **Add**. > [accordion] Each person connects their own account > > Each user signs in to the app with their credentials > > > [!NOTE] > > > > The **Server URL** and connection method - including your own OAuth app, if provided - can't be changed after you create the connector. > > 1. Navigate to **AI Hub > Connectors**. > 2. Click **Add connector**. > 3. Fill in the following: > * **Name** - Enter a name for the connector. This is what displays to users throughout the Omni UI. > * **Server URL** - Enter the server URL. > * **How do people connect?** - Select **Each person connects their own account**. > 4. **Optional**: Set **Advanced** options: > * **Transport** - Streamable HTTP (most servers) or SSE (older servers). Streamable HTTP is the default. > * **OAuth app** and **OAuth endpoints** - Bring your own OAuth client and endpoints for a sign-in connector that doesn't support dynamic registration. Leave these blank and Omni registers an app and discovers the endpoints for you. > 5. When finished, click **Add**. > > To use the connector, each user - including Organization Admins - must first [connect their own account](#connect-your-account). > [accordion] One shared API key for everyone > > All tool calls use a single API key > > > [!NOTE] > > > > The **Server URL** and connection method can't be changed after you create the connector. You can update your API key if needed. > > 1. Navigate to **AI Hub > Connectors**. > 2. Click **Add connector**. > 3. Fill in the following: > * **Name** - Enter a name for the connector. This is what displays to users throughout the Omni UI. > * **Server URL** - Enter the server URL. > * **How do people connect?** - Select **One shared API key for everyone**. > * **API key** - Paste your API key. > 4. **Optional**: Set **Advanced** options: > * **Transport** - Streamable HTTP (most servers) or SSE (older servers). Streamable HTTP is the default. > * **Header** - A custom header to send the API key under, if the server doesn't expect the default. > 5. When finished, click **Add**. > [!TIP] > > Some connectors work better when the Omni Agent knows more about them. After you add a connector, add details about it to the agent's [global context](https://docs.omni.co/ai/settings/global-context.md), such as what your team uses the app for and when the agent should use its tools. ## Connecting your own account If a connector uses the **Each person connects their own account** connection method, users must connect their individual accounts before they can use the connector. They can do this from a few places in Omni: * **The connector menu in the [Omni Agent](https://docs.omni.co/ai/chat.md)**. Click the icon, then **Connectors**, and then **Connect** next to the connector. * **From their user settings**. Connectors are listed in the [Connected services](https://docs.omni.co/administration/users/your-account.md#connected-services) section. Organization Admins can also connect from the **Connectors** page in the AI Hub. If an admin hasn't connected their own account to a connector yet, they can click **Connect your account** in the connector's row. To disconnect, see [Connected services](https://docs.omni.co/administration/users/your-account.md#connected-services) in Your Account. ## Editing connectors 1. Navigate to **AI Hub > Connectors**. 2. In the connector's row, click the icon, then **Edit**. 3. Some settings are read-only once the connector is created. Those you can update depend on the connector's [connection method](#connection-methods): > [accordion] No account needed > > * **Name** - The display name of the connector. > * **How people connect** - Select **Each person connects their own account** or **One shared API key for everyone**. If you select **One shared API key for everyone**, paste the API key in the **API key** field that displays. > > > [!NOTE] > > > > The connection method can't be changed again after you save. > * **Transport** - Streamable HTTP (most servers) or SSE (older servers). Streamable HTTP is the default. > * **Tool-call timeout** - How long Omni waits for one tool call to this server before giving up. This only applies to tool calls, not to health checks or tool discovery. > [accordion] Each person connects their own account > > * **Name** - The display name of the connector. > * **Transport** - Streamable HTTP (most servers) or SSE (older servers). Streamable HTTP is the default. > * **Tool-call timeout** - How long Omni waits for one tool call to this server before giving up. This only applies to tool calls, not to health checks or tool discovery. > [accordion] One shared API key for everyone > > * **Name** - The display name of the connector. > * **Replace API key** - Paste a new API key. > * **Transport** - Streamable HTTP (most servers) or SSE (older servers). Streamable HTTP is the default. > * **Header** - A custom header to send the API key under, if the server doesn't expect the default. > * **Tool-call timeout** - How long Omni waits for one tool call to this server before giving up. This only applies to tool calls, not to health checks or tool discovery. 4. Click **Save**. ## Removing connectors > [!WARNING] > > Removing a connector immediately stops its tools for everyone in the organization. If the connector uses **Each person connects their own account**, removing it means everyone who connected their account needs to re-connect if you re-add the connector. 1. Navigate to **AI Hub > Connectors**. 2. In the connector's row, click the icon, then **Remove**. 3. When prompted, confirm the removal. ## Limitations * Custom AI connectors aren't currently available for [Azure deployments](https://docs.omni.co/administration/security/cloud-regions.md). * Embedded instances can only use connectors that connect with a shared API key. Before you add a shared-key connector, see [Connectors in embedded sessions](#embed). ## Next steps * [Global context](https://docs.omni.co/ai/settings/global-context.md) - Add details about your connectors to the Omni Agent's global context This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.