Custom MCP
Pointing Emerge.ai at a remote MCP server so it appears as a connector alongside the built-in ones, what the panel needs, and what stays your responsibility once it is connected.
Not every tool a team uses is on Emerge.ai's list of connectors. Add Custom MCP lets you point Emerge.ai at a remote MCP server of your own.
MCP — Model Context Protocol — is a standard way for a service to offer its tools to an AI assistant. Point Emerge.ai at a server that speaks it and it appears alongside the built-in connectors, with its own Discovered Tools list on its detail page.
What it is, and what it isn't
A custom connector is a pointer. The panel asks for the address of a server that already exists and is reachable over HTTPS, and Emerge.ai connects to it. Emerge.ai hosts nothing for you.
Understand that before you plan around it. Add Custom MCP moves the work of building an integration outside Emerge.ai; it does not remove it. Something still has to run at that address, and somebody still has to keep it running. If nobody on your team owns that server, a custom connector will not get you the integration you want.
What the panel needs
Two fields do the ordinary work: a Name for the connector as you want it to appear, and the Remote MCP server URL.
Sign-in is OAuth only — the standard where you approve access on the provider's own site instead of handing Emerge.ai a password.
An Advanced settings group holds the OAuth details: client ID, client secret, authorization URL, token URL, and scopes. You only need them when the server does not publish those details automatically. Leave the group alone otherwise, and get the values from whoever runs the server.
There is no field for a static API key and no field for a custom header. That matters when you are choosing what to connect. A service that signs in with a fixed key or a custom header cannot be pointed at directly — its credentials would have to live inside the MCP server, not in Emerge.ai.
For the click-by-click version, see Set up a connector.
Its tools are discovered, not configured
A custom connector's tools come from the server itself: once connected, its detail page lists them as Discovered Tools, and the connector is usable in chat only after that discovery succeeds.
The per-tool Always allow / Ask each time switches described in Set up a connector are a built-in-connector feature today — the product notes that tool permissions for a custom connector become available once it can be provisioned. Treat a custom connector's tools as governed by the server you point Emerge.ai at.
Definitions
- Custom connector — a connector you add yourself with Add Custom MCP, pointing at a remote MCP server instead of an app Emerge.ai already lists.
- Remote MCP server — a server, reachable over HTTPS, that exposes tools Emerge.ai can call. It runs outside Emerge.ai and is operated by whoever set it up.
- Advanced settings — the optional OAuth group in the Add Custom MCP panel, used only when the server does not publish its OAuth details automatically.
Related topics
- What you can connect — the connectors Emerge.ai already provides
- Set up a connector — the steps, including the custom panel
- Connectors overview — what a connector is, and where connectors live
- Privacy & Security — what Emerge.ai does with what it reads
Set up a connector
Connect a provider app to your Cloud, decide which of its tools Emerge.ai may use without asking, and enable it in a chat.
Models & Auto-Selector
Choosing a specific model or letting Auto pick one, the groups the model picker presents, why you might choose Emerge.ai's own hosted models, and the models you can download for offline use on mobile.
