---
title: "Connect Codex"
description: "Add the chart-scoped SeatLayer Designer MCP Preview to Codex, authenticate with OAuth, and verify the connection safely."
---

Codex can connect directly to a remote Streamable HTTP MCP server. The ChatGPT
desktop app, Codex CLI, and Codex IDE extension share MCP configuration on the
same Codex host. See OpenAI's current
[MCP documentation](https://learn.chatgpt.com/docs/extend/mcp).

<Aside type="caution" title="Use the dashboard URL">
  Copy the Designer MCP URL from **Dashboard → Agent connections** in the
  correct SeatLayer environment. Do not copy a development URL into production.
</Aside>

## Desktop app or IDE

1. Open **Settings → MCP servers**.
2. Select **Add server**.
3. Name it `SeatLayer Designer`, choose **Streamable HTTP**, and paste the URL
   shown by SeatLayer.
4. Save and restart the app or extension.
5. Select **Authenticate** and approve the intended workspace and chart.
6. Use `/mcp` to confirm that the server is connected.

## CLI

Add the remote server to `~/.codex/config.toml` or the trusted project's
`.codex/config.toml`:

```toml
[mcp_servers.seatlayer_designer]
url = "YOUR_SEATLAYER_MCP_URL"
auth = "oauth"
default_tools_approval_mode = "writes"
```

Then run `codex mcp login seatlayer_designer` to authenticate and
`codex mcp list` to verify it. OpenAI documents the current configuration keys
and CLI commands in its [MCP guide](https://learn.chatgpt.com/docs/extend/mcp).
Do not place a SeatLayer secret API key in a prompt or configuration file.

## Verify safely

Start with a read and capability check:

```text
Use the SeatLayer Designer connector.
Call get_capabilities first, state the authorized chart scope, and summarize
the current chart. Do not mutate, approve, or publish anything.
```

Then follow the staged workflow in [Designer MCP](/agents/designer-mcp). Any
write must use the current expected revision, and publish still requires a
human-approved unchanged revision.