Set up an MCP connection

Prev Next

Who can use this: Tech admin

Available on: Any Gong plan

Ideal for: Sales manager, RevOps

Coming soon:

Gong MCP client and server are coming soon

Set up an MCP connection

Use MCP connections to connect Gong to external MCP servers.

To define an MCP connection:

  1. In the left sidebar, click Admin center.

  2. In the Settings tab, click MCP connections, under Ecosystem.

  3. Click New connection, and select New connection.

  4. Enter a name for the MCP connection. The name is displayed in Gong when choosing the MCP server you want to connect to.

  5. Enter the MCP server URL.

Set the authentication requirements:

  1. Toggle on Authentication required if the MCP server requires credentials to set up the connection.

  2. For servers that authenticate with OAuth 2.0 with pre-configured credentials: Expand Client access and provide:

    • Client_id

    • Client secret

  3. Under Who can use this connection, select whether the connection will allow:

    • Shared access: A single token is used for all users to connect to the external MCP server and individual users don’t need to sign in separately.

    • Personal access: The user who sets up the initial connection authorizes the connection at a company level, and each individual user must then authorize their own account separately.

  4. Click Connect.

  5. Complete authentication if prompted.

FAQs

When should I choose shared access and when should I choose personal access?

Use shared access when:

  • The connection is used for backend or system-level integrations

  • The external MCP server provides shared resources

  • Individual user identity does not affect the data returned

  • The use case involves automation, reporting agents, or pipelines

Shared access is best suited for machine-to-machine scenarios where a single connection represents the organization.

Use personal access when:

  • The connection is used in interactive tools such as AI clients

  • The external MCP server enforces per-user permissions

  • The data returned depends on the individual user’s access

Common examples include:

  • Chat-based AI tools such as Claude, ChatGPT, or Copilot

  • Systems with user-specific data such as private Jira boards or personal Notion pages