Who can use this: Tech admin
Available on: Any Gong plan
The Gong MCP server exposes Gong insights to external AI systems.
External AI clients, such as chat-based assistants or embedded AI tools, can query the Gong MCP server to retrieve AI-generated insights based on Gong activity data.
Available connectors
Gong tech admins can enable the Gong MCP server for:
Once these have been enabled, team members can add the Gong connector in their AI tool.
How the Gong MCP server works
External MCP clients send requests to the Gong MCP server using JSON-RPC.
The MCP server processes each request by analyzing Gong activity data and returns AI-generated insights in a structured response.
MCP server tools
The MCP server exposes the following tools:
ask_account: Answers natural language questions about a single account
ask_deal: Answers natural language questions about a single deal
generate_brief: Generates a structured, multi-category summary (brief) for an account, deal, or contact, including insights such as key themes, stakeholders, risks, and next steps
Each tool analyzes calls and emails within a defined time range and returns AI-generated insights.
Include sources
Each tool has an includeSources parameter which by default is set to false. When true, the tools return a link to the calls and emails used to generate the answer. This toggle is turned on automatically when the prompt includes a request to return the sources used to generate the answer. When false, only the text is returned.
ask_account
The ask_account tool answers a specific question about a single account.
It analyzes Gong activities linked to the account and returns a focused, natural-language response. The response highlights patterns such as objections, risks, stakeholder concerns, and competitive mentions based on recent interactions.
Use this tool when the external AI agent needs a targeted insight about an account. This includes scenarios such as understanding customer concerns, identifying risks, or preparing for an account review.
This tool works best when the external AI agent asks a concise, insight-oriented question. It returns a single synthesized answer and does not provide raw activity data.
Do not use this tool for broad summaries or multi-topic overviews. When a structured summary across multiple categories is required, use the generate_brief tool.
ask_deal
The ask_deal tool answers natural language questions about a single CRM deal.
It analyzes calls and emails linked to the deal and returns a focused, natural-language response. The response highlights patterns such as deal blockers, stakeholder concerns, risks, and agreed next steps based on recent interactions.
Use this tool when the external AI agent needs a targeted insight about a deal. This includes scenarios such as understanding what is preventing a deal from closing, tracking deal progress, or reviewing recent commitments.
This tool works best when the external AI agent asks a concise, insight-oriented question. It returns a single synthesized answer and does not provide raw activity data.
Do not use this tool for structured summaries or multi-topic overviews. When a broader, multi-category view of the deal is required, use the generate_brief tool.
generate_brief
The generate_brief tool generates structured summaries for CRM accounts, deals, or contacts.
It analyzes calls and emails linked to the account or deal and returns a comprehensive summary organized into predefined categories such as key themes, stakeholders, risks, and next steps.
Note:
For contacts, you must provide the CRM contact ID is required. For accounts and deals, name lookup is supported.
Use this tool when the external AI agent needs a broad overview rather than a single answer. This includes scenarios such as account reviews, deal reviews, executive briefings, or handoffs.
This tool works best when a structured summary across multiple topics is required. It returns a single, organized output based on a predefined brief and does not accept a question.
Do not use this tool when the external AI agent needs a specific answer to a specific question. For targeted insights, use the ask_account or ask_deal tools.
Data and behavior
The MCP server returns AI-generated insights based on Gong data.
It analyzes calls and emails within a defined time range and returns summarized responses designed for decision-making.
Raw data such as call transcripts, message bodies, and activity lists is not returned.
The MCP server is read-only and does not create, update, or delete data in Gong or the CRM. Private calls are excluded from all results.
Registration types
Registration type determines how the integration is made available, through a public marketplace or as a private integration with specific credentials.
Automatic: No credentials required. The client either connects from a public marketplace or app store, such as the ChatGPT Apps directory or the Microsoft Copilot marketplace, or auto-registers with Gong against an approved redirect URI. Team members connect through a standard Gong sign-in flow, and may connect using a personal AI client account as well as a company one.
Manual: A private integration not listed in any marketplace. Gong generates a client ID and client secret which the admin installs in the connecting AI client, for example an enterprise Claude or ChatGPT console. Only that credential can connect, so marketplace connections and personal accounts are blocked for the company.
Authorization context
Authorization context determines what data a connected client can access. It is configured independently from the registration type, so any combination of the two is possible.
The MCP server supports two authorization context options:
Personal access: Access limited to the authenticated team member's own data. Each team member connects with their own Gong credentials and can only access data they are already permitted to view in Gong.
Shared access: Organization-wide access through a single authorized token. All team members connecting through the integration see the same data, regardless of their individual identity.
Because each integration is a self-contained configuration, you can create more than one integration for the same AI client. For example, you could set up two separate ChatGPT integrations, one with shared access for a company-wide view, and one with personal access for individual team members, and each would behave independently.