MCP integration

Connect Codex to PrompTessor MCP

Set up PrompTessor in Codex settings or with the Codex command line, then complete OAuth authorization.

Verified against the implementation · 2026-09-09

Codex, IDE extension, and ChatGPT desktop app

Codex, the Codex IDE extension, and the ChatGPT desktop app share MCP configuration for the same Codex host. Add PrompTessor once and authenticate with OAuth; a PrompTessor REST API key is not required.

No terminal setup required

Use this method in the ChatGPT desktop app or Codex IDE extension. Both surfaces use the same Codex MCP configuration, so you do not need to repeat the setup in each one.

Open Settings > MCP servers and select Add server.

Name: PrompTessor
Server URL: https://mcp.promptessor.com/mcp
Transport: Streamable HTTP
Authentication: OAuth

1
Open MCP settings
In the ChatGPT desktop app, open Settings > MCP servers. In the Codex IDE extension, open the gear menu and select MCP servers.
2
Add a server
Select Add server, enter PrompTessor as the name, and choose Streamable HTTP as the transport.
3
Enter the server URL
Paste https://mcp.promptessor.com/mcp in the URL field, then save the server.

Codex custom MCP settings with PrompTessor name, Streamable HTTP transport, and server URL — PrompTessor custom MCP configuration. Leave Bearer token, headers, and environment-variable fields empty, then select Save.

4
Restart the client
Restart the ChatGPT desktop app or select Restart extension in the IDE so the new MCP server is loaded.
5
Authenticate
Select Authenticate beside PrompTessor, then sign in to PrompTessor and approve the OAuth connection in your browser.

Codex MCP server list showing the Authenticate button for PrompTessor — After saving, find PrompTessor in the server list and select Authenticate to start OAuth.

6
Review and allow access
Confirm that the page says Connect to PrompTessor, review the requested permissions, and select Allow access. Only the permissions shown on this screen are granted.

PrompTessor OAuth consent screen for Codex showing requested access and the Allow access button — Review the requested PrompTessor permissions before selecting Allow access. You can disconnect this Codex client later from PrompTessor Settings.

7
Verify the connection
Open /mcp to confirm that PrompTessor and its tools are available, then ask Codex to check your current PrompTessor usage.

Authenticate is missing after reinstall

Removing and adding PrompTessor again may leave the previous OAuth session stored on the shared Codex host. The client can therefore treat the server as already authenticated even when the saved refresh token has expired or been revoked. Restarting the app or extension does not clear that stored session.

codex mcp logout promptessor
codex mcp login promptessor

1
Confirm the saved server
Run codex mcp list and confirm that promptessor is present. Use the npx fallback when the codex command is not installed globally.
2
Clear only the old PrompTessor session
Run codex mcp logout promptessor. This clears the saved OAuth session for PrompTessor without removing the MCP server configuration.
3
Authenticate again
Run codex mcp login promptessor and complete sign-in in the browser. Alternatively, restart the app or extension after logout and select Authenticate when it reappears.
4
Verify the repaired connection
Open /mcp or run codex mcp list, then ask Codex to call get_usage. If login still fails, copy the exact OAuth error and send it to PrompTessor Support.

Do not delete the complete OAuth credential store

Use mcp logout promptessor instead of deleting Codex OAuth credential files manually. The shared credential store can contain sessions for other MCP servers.

Install Codex CLI before using codex commands

Installing the Codex IDE extension does not guarantee that the codex command is available in your terminal PATH. If “codex is not recognized” or “command not found” appears, install the CLI, reopen the terminal, and verify codex --version before continuing.

npm install --global @openai/codex@latest
codex --version

Server name: promptessor
Server URL: https://mcp.promptessor.com/mcp
Transport: Streamable HTTP
Authentication: OAuth
CLI install: npm install --global @openai/codex@latest
Config file: ~/.codex/config.toml or .codex/config.toml

1
Check Node.js and npm
Run node --version and npm --version. Install a current Node.js release first if either command is unavailable.
2
Install Codex CLI
Run npm install --global @openai/codex@latest. When installation finishes, close and reopen the terminal. Restart VS Code as well when using its integrated terminal.
3
Verify the command
Run codex --version before any MCP command. On Windows, if codex is still not recognized, run npm config get prefix and add that returned directory—commonly %APPDATA%\npm—to your User PATH, then restart VS Code.
4
Add the server
Run codex mcp add promptessor --url https://mcp.promptessor.com/mcp. If you cannot install globally, use the npx fallback shown above.
5
Authenticate
Run codex mcp login promptessor. Complete PrompTessor sign-in and OAuth consent in the browser.
6
Check server status
Run codex mcp list. In the Codex terminal UI, enter /mcp to confirm that PrompTessor is enabled and its tools are visible.
7
Verify with a safe call
Ask Codex to use PrompTessor to check your current usage. It should call get_usage without spending an AI credit.

Official Codex CLI guideInstall, sign in, and run Codex from a terminal.Official OpenAI MCP documentationConfiguration, OAuth, tool filtering, and approval options.