Connect ChatGPT to AssinaJá
The AssinaJá side is identical to Claude.ai: one API key, one two-step authorisation. What differs is the ChatGPT side, which needs one extra step — custom MCP connectors live behind developer mode, and it ships off.
Before you start
Section titled “Before you start”- In AssinaJá: be the organisation’s Owner. Only the Owner creates API keys, and the authorisation only lists keys of organisations you own.
- In ChatGPT: a Pro, Plus, Business, Enterprise or Education account, on the web. On a company account, an administrator has to allow custom connectors before the setting appears to you.
1. Create the key in AssinaJá
Section titled “1. Create the key in AssinaJá”Under Settings → Developers → API Keys, create a key with a name that identifies the connection — ChatGPT. The steps are in Create an API key, and the permissions to tick for a first connection under Scopes.
2. Turn on developer mode
Section titled “2. Turn on developer mode”The steps below are from the ChatGPT interface in August 2026. This is the area whose names change most — the official guide is at developers.openai.com.
- Open ChatGPT at chatgpt.com.
- Go to Settings → Security and login.
- Turn on Developer mode.
On a company account, if the option is absent, an administrator releases it in the workspace settings, under connector and developer-mode permissions.
3. Add the connector
Section titled “3. Add the connector”Still in settings, create a developer-mode app for a remote MCP server. You are asked for:
| Field | Value |
|---|---|
| Server URL | https://app.assinaja.pt/mcp |
| Authentication | OAuth |
| Name | your choice — AssinaJá |
AssinaJá supports streamable HTTP, which is the transport ChatGPT uses for remote servers.
Once created, the connector has its own page where you can turn individual tools on and off, and re-read the server to pick up new ones.
4. Authorise
Section titled “4. Authorise”ChatGPT opens the AssinaJá authorisation page — Connect ChatGPT to AssinaJá. Sign in if you have not already.
Step 1 of 2 — choose the API Key to use
From the list (— select an API Key —), pick the key you created. Continue.
Step 2 of 2 — confirm the access you are granting
Review what the key allows:
By confirming, it is recorded that you authorized ChatGPT to use this key. You can cut the access at any time by suspending or deleting the key under Settings → API Keys.
Authorize access, and you are back in ChatGPT. What each step checks, and what is recorded, is in Authorise and revoke an assistant.
5. Confirm it works
Section titled “5. Confirm it works”In a conversation, enable the developer-mode tool and ask:
List the templates available in AssinaJá.
If you get your real templates, you are connected. If you get a generic answer, the connector is not being used in that conversation — in ChatGPT, developer-mode tools are selected per conversation rather than staying on.
That is the practical difference from Claude.ai, and the most common cause of “it worked yesterday”.
6. Cut access
Section titled “6. Cut access”As with Claude.ai: from the key’s side, not the assistant’s.
Settings → Developers → API Keys → Suspend. Access drops on the next request, and you can reactivate later. Deleting is permanent. See Authorise and revoke an assistant.
If it goes wrong
Section titled “If it goes wrong”| What you see | What to do |
|---|---|
| No custom-connector option | Developer mode is off, or the administrator does not allow it in this workspace. |
| You have no active API Key. | The key does not exist, is suspended, or expired — or you are not the Owner of the key’s organisation. |
| The connector connects but no call works | Authentication was left as No authentication. Redo it with OAuth. |
| Tools do not appear in the conversation | Enable the developer-mode tool in that conversation. |
| Tools are missing from the list | The key lacks that permission — see The assistant only sees what the key allows. |
| Nothing changes after editing the key’s permissions | Re-read the server on the connector’s page so ChatGPT refreshes the list. |