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

# Connect a remote MCP server

> Add a connector, bring it into a Space, and authorize it there.

A [Connector](/concepts/connectors) links the Hub to a **remote MCP server** so your
skills and agents can use that service's tools. Authorization uses the standard
**OAuth** flow in your browser, and the Hub calls the server directly from your machine.

<Warning>
  **If you used connectors before this release, you'll sign in again.** Connectors are
  now authorized per [Space](/concepts/spaces) rather than once globally. Existing
  connections are not carried over — your connectors are still listed, but each Space
  needs its own authorization. This is a one-time step per Space.
</Warning>

## Find a connector

Two ways in:

* **The Connector Directory** — a curated, in-app list of services that are known to
  work. Browse it, and add what you need to your connector list.
* **Any remote MCP server, by URL** — paste an HTTP MCP server's URL, or its JSON
  config, to add something the directory doesn't carry.
  [mcpservers.org](https://mcpservers.org/remote-mcp-servers) is a good index of
  what's out there.

Adding a connector here only makes it *available*. It doesn't sign you in.

## Connect it in a Space

<Steps>
  <Step title="Add the connector to your list">
    From the Connector Directory, or by URL for a custom server.
  </Step>

  <Step title="Add it to a Space">
    Open the Space's connectors and add it from your list. It lands **added but not
    authorized** — present, with no tools loaded yet.
  </Step>

  <Step title="Authorize">
    Choose **Authenticate**. Your browser opens the service's consent screen; approve
    it. Some services need an OAuth app you register yourself first — see
    [Bring your own OAuth app](/connectors/oauth-apps).
  </Step>

  <Step title="Choose the tools">
    Once authorized, the connector's tools load with descriptions. Enable them
    individually, or all at once. Only enabled tools are offered to agent runs.
  </Step>
</Steps>

Repeat step 2 onward for each Space that needs it. Each Space holds its own
authorization, so different Spaces can sign into different accounts on the same
service.

## Turning a connector off

Disabling a connector in a Space keeps it authorized and configured but takes it out of
agent runs — useful when you want it available but not in play.

**Removing** it from the Space deletes that Space's credentials for it. Other Spaces
keep theirs.

## Use the tools

Once authorized and enabled, the server's tools are available by name to that Space's
[skills](/masters/skill) and agents. The Hub keeps the tool list in sync, and you can
re-authorize, disable, or remove a connector at any time.

<Note>
  A worked example with a real connector (and screenshots) is being added to this
  page.
</Note>