> ## 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.

# Connections and agent access

> Connect an external agent to MTS or assign Gatetrix Access tools to an embedded teammate.

There are two directions of connection: an external agent can use MTS tools, and
an embedded MTS teammate can use tools from Gatetrix Access.

## Connect an external agent to MTS

This route works without the built-in Codex integration or signing in to ChatGPT
in MTS. Use an MCP-compatible coding agent with its own runtime and provider.
MTS supplies project records and workflow tools; the external agent performs
the coding work. Its own subscription or API billing is managed in that client.

1. Open **Agent connections** at the bottom of the sidebar.
2. Enable **Allow agent access** and the MCP transport your client will use.
3. Choose **New connection**, name it, and select the projects it may access.
4. Choose its **Allowed actions**. Grant only the actions needed for the intended work.
5. Create the credential and use the copied MCP configuration or setup command
   in your client. Keep the token out of chat and source control.
6. Reconnect the client and ask it to list the projects available to the connection.

**Selected projects** limits the connection to those projects. Selecting every
current project is different from granting **All current and future projects**.

For local HTTP, use **This Mac** and the loopback endpoint shown by MTS. For
another device, use an authenticated, encrypted private VPN between the devices
and connect through the Mac's VPN address with the configured port and path.
Plain HTTP alone does not protect traffic or bearer tokens in transit, even on
a network you trust. Do not expose the endpoint directly to the public internet.
Keep MTS open while using its tools.

For a first check, ask the agent to list its available projects and read one
selected task. Then ask it to follow that task's workflow. Keep its project and
action permissions narrow; an external coding runtime's filesystem and network
permissions are configured in that runtime.

## Give an MTS teammate Access tools

Create a dedicated profile in [Access](/access/quickstart), configure its
connections and permissions, and enable local MCP HTTP. In the teammate's chat,
open the link button (**Member integrations**), choose **Add connection**, and
enter that endpoint and profile credential. Enable **Assigned** for the connection
and use **Test** to confirm tool discovery.

The teammate can discover tools on its assigned connections. Service credentials
stay in Access; MTS stores the Access connection token in Keychain. Access
continues to enforce its profile permissions and approvals.

When a call requires approval, decide in Access and have the agent check the
saved request result. Repeating the original operation is not the way to retrieve
an approved result.

## MTS actions and execution access

Allowed MTS actions control project records: reading, suggesting, editing,
organizing, working on tasks, accepting suggestions, restoring/trashing content,
and managing project configuration.

A teammate's **Execution access** is separate. It controls the coding runtime's
filesystem, command, and network access. **Full access** can include files outside
the project; choose it only when that scope matches your intent. Project
permissions do not themselves constrain everything a coding runtime can do.

After an upgrade or connection change, refresh an external client's discovery.
Start a new MTS conversation when new integration tools are missing from an
existing conversation's saved tool catalogue.


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