Use Ondes PRO with ChatGPT

Developer mode is an OpenAI beta feature, available on the web. Its availability — and, more importantly, whether ChatGPT can write rather than just read through a custom connector — varies by plan: OpenAI documents it for Business, Enterprise and Edu, and public information is inconsistent for individual plans. In an organisation, an admin often has to allow it first.

Our advice: check OpenAI’s help page for what your plan allows when you read this, rather than trusting a table that will have aged. On the Ondes PRO side nothing changes between ChatGPT plans: the server exposes the same eight tools and enforces the same permissions.

ChatGPT is comfortable with operational questions that span several modules: preparing a meeting, summarising the store’s week, spotting overdue invoices, finding who owns an account. For these it readily goes through the ondes_assistant tool, which queries the Ondes PRO built-in assistant — the one that knows your modules and pulls the answers together.

For writes, two safeguards stack up. On the ChatGPT side, an action that changes data goes through a confirmation in the conversation, when your plan allows writes at all. On the Ondes PRO side, the built-in assistant presents its changes as a plan, executed only after you approve; and a direct API call goes to test mode on resources that have one.

First, that developer mode is on and the connector is selected in the conversation: a custom connector is not used until it is offered to ChatGPT. Then, that your Ondes PRO role includes the API & Developers module, without which the connection is refused.

Finally, if ChatGPT ignores a module, it is almost always because your role does not open it: the server never documents a resource to the assistant that would answer “access denied”. Ask your workspace admin to adjust your role, then reconnect.