How to Configure Connectors

GLChat Deep Research can use various data sources to enrich results. Connectors allow the system to search and retrieve information from external services.

In the future, we will support two main ways to configure connectors for the GLChat DeepResearch Internal profile:

GL Connector

Internal connector service with pre-built integrations (e.g., GitHub, Gmail, Calendar, Google Drive). See GL Connectors Quickstart.

How it works: The user authenticates with the provider (e.g., Google Drive) via GL Connector. Once connected, GLChat can access the user's data from that provider for research and other features.

MCP (Model Context Protocol)

Custom tools and data sources by connecting to MCP-compliant servers.

How it works: The user creates their own MCP server (exposing the data or logic they want to search) and connects it to GLChat. During Internal research, the system invokes the user's MCP server as one of its tools to access their custom data sources. For how to create an MCP server, see MCP Server Hello World.

MCP as a Custom Pipeline

The MCP server can be a pipeline that you build (e.g., a RAG pipeline). You create the pipeline, expose it as an MCP server, and connect it to GLChat. See Your First RAG Pipeline for how to create a pipeline.

MCP as a Prebuilt GLChat Pipeline

The MCP server can be a prebuilt pipeline from GLChat (e.g., RAG). In this case, you do not need to build the pipeline yourself. You simply configure the connection to the existing GLChat pipeline, and the Internal profile will use it as one of its tools during research.

Configuration guides will be added when these features are ready.