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

# Connect your first agent

> Install Gatetrix Access, create an agent profile, and connect an MCP client using configuration generated by the app.

You need an Apple silicon Mac running macOS 13 or newer and an agent application
that supports MCP. Access uses a separate background core; its window can be
closed while the app stays running. Read [Interfaces](/access/interfaces#background-service-and-closing-the-app)
for the difference between closing the window and quitting.

<Steps>
  <Step title="Install Access">
    Download [Gatetrix Access](https://gatetrix.ai/download/), open the disk image,
    and install the app in Applications. Open Access.
  </Step>

  <Step title="Create an agent profile">
    Open **Configuration** and create an **Agent Profile** for this agent or task.
    Give it a recognizable name, such as “Research assistant.” Select that profile
    before editing its connections and permissions.
  </Step>

  <Step title="Enable an MCP transport">
    Under **Access Channels → MCP**, enable MCP and the transport your client uses.
    Use stdio for a local process connection, or HTTP bound to `127.0.0.1` for a
    client on this Mac. Use the endpoint shown in Access.
  </Step>

  <Step title="Create the credential and connect your client">
    Open **MCP Credentials** for the selected profile and create a credential.
    Use the app's generated client configuration for your chosen transport.
    Put the credential in your MCP client's credential configuration, not in a
    conversation or project source file. Reconnect the client after configuration.
  </Step>

  <Step title="Add one useful data source">
    Open **Connectors** or **Folders** for the profile and grant the source you
    want this agent to use. For an online service, follow
    [Set up a connector with an agent](/access/connectors).
    Start with the smallest read scope that answers your question.
  </Step>

  <Step title="Verify access">
    Ask the agent to discover its tools and perform one bounded read from the
    approved source. Check that the result is useful and inspect **Audit Log**.
    If approval is required, decide in Access and have the agent retrieve the
    saved result.
  </Step>
</Steps>

## What success looks like

The agent can discover and use the intended tool through its profile. It has
not needed the connected service's API key in chat. The activity appears in
Access under the correct profile.

If tools are missing, check the selected profile, credential, transport, and
connection grants. An upstream MCP server connection also needs tool discovery
and review. See [Troubleshooting](/support) for the next checks.


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