1
Open Connectors
In the dashboard, open Connectors. The first section lists servers the vendor runs; below it,
Managed by BoostGPT lists the connectors we maintain.
2
Connect OpenAI
Find OpenAI under Make images, video and voice for the site and press Connect. Paste an API key
from platform.openai.com/api-keys.The key is checked against OpenAI before anything is saved. If it is refused you will see OpenAI’s own words
— “Incorrect API key provided” — and nothing is stored, so you never end up with a connection that looks
fine and fails on its first real call.
3
Give it to an agent
Open an agent, go to its Connectors tab, and add OpenAI. No credential is asked for again: the
connection belongs to the workspace, and the agent is simply allowed to use it.
4
Ask for something
In the chat, try:The agent picks
generate_image, calls it, and the result comes back in the conversation.A connection that exists is one that works. If Connect refused your key, that is the check doing its job —
fix the key and press Connect again.
Connecting a server you do not have a key for
Several services run their own MCP server, and those need no key at all — you sign in and approve access, and BoostGPT introduces itself to the server automatically. MetaManager is a good first one: it scores how a page presents itself, which is immediately useful on a Site you have just built. See remote MCP servers for the full list.What to do next
Browse the connectors
Everything BoostGPT maintains, grouped by what it is for.
Bring your own API
Turn an OpenAPI specification or a Postman collection into a connector.