I have an MCP server published to ChatGPT Enterprise and working in normal chats. I have also added the MCP app to a custom GPT. The issue is, when I now try use the MCP tooling from the custom GPT, it can’t find the tooling. Is there any additional steps that are needed for gotcha’s that I haven’t found?
Hey @cor, yeah this one can be confusing. Apps in custom GPTs is currently rolling out in beta to Business and Enterprise/Edu accounts over time. If you don’t see it yet, it should show up as the rollout reaches your account, so it’s worth checking back in a few days. We can’t share an exact ETA for individual accounts right now.
For more information you can visit: Apps in custom GPTs (beta)
@OpenAI_Support I am on Enterprise and have apps enabled in custom GPTs. I’m able to connect the app to the custom GPT, but it isn’t recognising it during the chat. I’m having the same issue when try to use third-party apps such as Slack - it’s just not able to use the app even though its connected.
If your MCP tools work in normal Enterprise chats but not inside a Custom GPT, it’s usually a permissions or configuration issue. Make sure the MCP app is explicitly enabled in the Custom GPT’s Actions/Tools settings, and that the GPT has access to the same workspace and authentication scope as your main chat. Sometimes re-adding the MCP app and republishing the GPT fixes caching issues too.
Also double-check that the Custom GPT is set to use external tools and that no policy restrictions are blocking it.
Seemed to have fixed the issue by disabling developer mode.