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

# Connected Services

> Connect RelayHub to external tool servers so their tools work in your chats and agents

Connected Services let RelayHub reach OUT to an external tool server and use its
tools inside RelayHub. Once a service is connected, its tools become available to
the AI in chat, to your agents, and to agent chains, right alongside RelayHub's
own tools. This is the mirror image of [AI Assistant Connections](/integrations/mcp-connections):
there, an outside assistant connects IN to RelayHub; here, RelayHub connects OUT
to an outside tool server.

Both sides use the same open standard, the Model Context Protocol (MCP), so any
server that speaks MCP can be connected.

<Note>
  Connected Services are managed by administrators. You will find them under the
  platform navigation as **Connected Services** (admin only).
</Note>

## When to use it

Connect a service when a third party publishes an MCP tool server you want your
RelayHub AI to use, for example a vendor's search API, a data provider, or your
own internal tool server. After you connect it, you can just ask in chat and the
AI will call the external tool for you, no extra steps.

## Setting one up

1. Open **Connected Services** from the platform navigation.
2. Click **Add Service**.
3. Fill in:
   * **Name**: a label you will recognize (for example "RankPilot Tools").
   * **Server URL**: the server's MCP endpoint. It must be an `https` address.
   * **Authentication**: how RelayHub proves who it is to that server:
     * **None**: the server needs no credentials. Not sure? Pick this one. If the
       server requires a sign-in, RelayHub detects that and opens the sign-in
       window automatically when you connect.
     * **API key (Bearer)**: paste the key the server's owner gave you. RelayHub
       stores it encrypted and sends it as an `Authorization: Bearer` header. If
       the server wants a different header, such as `X-API-Key`, set the optional
       header name.
     * **OAuth**: sign-in based authentication. The first connect opens a popup
       where you sign in and approve access. RelayHub registers itself with the
       server automatically and stores the tokens encrypted.
4. Click **Add & Connect**. RelayHub connects to the server, reads the list of
   tools it offers, and makes them available. You will see the tool count and can
   expand the row to view each tool.

<Note>
  Some providers restrict automatic registration to specific apps. In that case the
  connect fails with a clear message from the provider. Ask the provider to allow
  RelayHub, or ask them for an OAuth client ID and enter it in the connector's
  settings. A client ID from the provider makes the sign-in work without automatic
  registration.
</Note>

## Using the tools

Once a service shows **Connected**, its tools are live everywhere the AI runs: chat,
agents, and agent chains. You do not select them manually. Just ask for what you
want ("look up the AI visibility for acme.com") and the AI decides when to call an
external tool, the same way it decides to search your files or run a skill.

Each external tool is named `mcp_<service>_<tool>` so it never clashes with a
RelayHub tool or with another service's tool.

**Skills can call these tools too.** Add the tool's name to the skill's
`platform_tools` list in the Skill Studio, and the skill can call the connected
service during its run. Expand a service's row on this screen to see the exact
tool names. The call runs with the same organization isolation and audit trail
as a call from chat.

## Managing a connection

On the Connected Services page you can:

* **Refresh**: re-read the server's tool list (use this after the server adds or
  removes tools).
* **Edit**: change the name, address, or authentication of an existing connection.
  You can rotate the API key (leave the field blank to keep the stored one) or
  switch the authentication method. A method change clears the stored credential
  and asks you to reconnect, so an old secret is never reused the wrong way.
* **Test**: check the connection without changing anything.
* **Disable / Enable**: turn a service off without deleting it. A disabled service's
  tools immediately stop being offered to the AI.
* **Remove**: delete the connection and its stored credentials.

## Security and privacy

* **Credentials are encrypted.** An API key or OAuth token is stored encrypted and
  is never shown back to you or written to logs.
* **Each connection is private to your organization.** One organization can never
  see or use another organization's connected services, credentials, or tools.
* **Only safe destinations are allowed.** RelayHub refuses a server address that
  points at an internal or private network, so a connection can never be used to
  reach something it should not.
* **Every external tool call is audited.** Administrators can review connected-service
  activity in the Audit Hub, the same as any other action.
* **Changing the address clears the key.** If you repoint a service at a different
  host, RelayHub clears the stored credential so the old key is never sent to the
  new address. You reconnect deliberately.

## Limits

* An `https` address is required; plain `http` and internal hosts are rejected.
* There are per-organization caps on how many services you can connect and how many
  tools RelayHub ingests from any one server, to keep the AI fast.
* If your administrator set an allowlist of permitted hosts, only those hosts can be
  connected.