What you can actually do with the Gong MCP
Gong's MCP lets an AI assistant ask about one account or deal at a time. Four/Four's MCP lets it work across every conversation, insight, account and CRM record you have, all at once.
We get asked a lot what people are doing with Gong's MCP connection in Claude, ChatGPT and Copilot. So here's what Gong's own documentation says you can reach, which tools it exposes, and where the limits are. Then we compare that with Four/Four.
Four/Four connects to Gong directly, so you can keep Gong and still get everything its MCP leaves out.
Gong: three question-answering tools, no raw data
Gong's MCP server is read-only and has three tools. Each one runs Gong's own AI over your calls and sends back a written answer.
| Tool | What it does |
|---|---|
ask_account |
Answers one plain-English question about one account |
ask_deal |
Answers one question about one CRM deal, using the calls linked to it |
generate_brief |
Writes a sectioned summary (themes, stakeholders, risks, next steps) for one account, deal or contact |
What that means in practice:
- No transcripts. Gong says plainly that "raw data such as call transcripts, message bodies, and activity lists is not returned." The only option,
includeSources, adds links to the calls used. - One record per question. Every tool is scoped to a single account, deal or contact. There's no tool for questions across your whole customer base, like "what are all our enterprise customers saying about onboarding?"
- Unpublished limits. The MCP docs give no default date range, maximum range or call cap. Gong's in-app Ask Anything feature, which looks like the same engine, draws on up to 60 calls for a deal or account, and up to 10 for a contact.
- Credits for every question. Each request uses Gong credits, and repeat questions use them again. Four questions about one account analyse the same calls four times. Briefs with more sections use more credits, because each section runs its own analysis.
- Gong data only. Answers come from what Gong recorded. Support tickets, chats and product-tracker issues aren't included.
To get full transcripts out of Gong, you have to go through its REST API or a community-built MCP server, with paging and rate limits to manage yourself. Or you can connect Gong to Four/Four.
Four/Four: conversations, insights, themes and CRM in one connection
Four/Four's MCP has 15 data tools, plus a help tool. They give an AI assistant your raw conversations plus everything Four/Four has already extracted from them, joined to your CRM and product tracker. Every search tool is paged, so the numbers below are per page, not a ceiling: an assistant asks for the next page to keep going.
| Group | Tools | What you get |
|---|---|---|
| Conversations | find_conversations |
Meetings, phone calls, support tickets, chat threads and email threads. Each with a summary and playback link, plus the full transcript when you ask for it |
| Insights | find_insights |
Unlimited. Each one cites its source conversation and transcript position, with the CRM account and contact where known. Typed as a job to be done, pain, impact, feature request, needs analysis, solution, support activity, effort, emotion, force or value, and most types have their own sub-categories to filter on |
| Themes | create_topic_model, find_topic_models, find_topics |
Groups any filtered set of insights into themes on the server. Each theme has an insight count, a link, and the customers, open deals and leads that mentioned it most recently. Insights can then be filtered by theme |
| CRM | find_accounts, find_contacts, find_leads, find_opportunities |
Firmographics, contacts, leads and opportunities. Opportunities filter by stage, amount, status or type; leads by source, stage or status |
| Product | find_issues |
Issues and ideas from connected product trackers, with project, status, type, quality and source links |
| Organise | create_label, attach_label, detach_label, modify_label, find_labels |
Personal or team labels you can attach to records and use as filters in other tools |
The conversation, insight and topic-model tools share the same filters: account and account labels, account owner and type, industry, company size, region, opportunity stage, status, type and amount (current or at the time of the conversation), lead source, stage and status, persona role and department, conversation type and purpose, and a fixed or relative date range ("last 2 weeks"). The CRM, issue and label tools have their own filters, and every tool takes a free-text search. Every result links back to the source in Four/Four.
Side by side
| Gong | Four/Four | |
|---|---|---|
| Tools | 3 | 15, plus help |
| Full transcripts | No | Yes |
| Scope of a question | One account, deal or contact | Any filtered slice of your whole customer base |
| Pre-extracted insights | No, AI answers only | Yes |
| Theme analysis across conversations | No | Yes, topic models built on the server |
| Sources | Gong calls and emails | Meetings, calls, support tickets, chats and emails, plus Gong calls |
| CRM joined in | Accounts and deals, as the scope of a question | Accounts, contacts, leads and opportunities, as filters and results |
| Product tracker | No | Yes |
| Write actions | None | Labels, topic models |
| Published limits | Yes, every question uses credits | None |
Already on Gong? Four/Four fills the gaps
You don't have to switch away from Gong. Four/Four connects to Gong and brings its calls into the same place as your other conversations. Your AI assistant then gets everything Gong's MCP leaves out, on the same calls.
| What Gong's MCP leaves out | What Four/Four adds to the same calls |
|---|---|
| No transcripts | Full transcripts through find_conversations |
| One account or deal per question, with credits used each time | One filtered query across every account |
| AI-written answers only, nothing to count or filter | Insights per page, already extracted and typed |
| No theme analysis across calls | Topic models with insight counts and linked customers and deals |
| No product tracker | Calls joined to accounts, contacts, opportunities and tracker issues |
| Gong calls and emails only | Gong calls combined with support tickets, chats and other meetings |
Gong keeps recording your calls. Four/Four is the layer your AI tools query to make sense of them.
Why the combined offer matters
Gong's MCP answers questions about one deal at a time. Four/Four lets an assistant reason across your whole customer base, because the hard work is already done before the question is asked.
- The analysis happens before the question. Four/Four has already extracted, typed and cited insights from every conversation. The assistant pulls pains or feature request insights , instead of asking Gong the same question account by account.
- Questions across accounts, not one at a time. "What are enterprise customers in financial services saying about onboarding this quarter?" is one filtered query. With Gong it means looping
ask_accountacross every account and paying credits each time. - Themes with counts and revenue attached. Topic models group thousands of insights into themes on the server. Each theme shows how many insights back it and which customers and open deals raised it. That's the evidence a roadmap or pricing decision needs.
- Every channel, not just recorded calls. Support tickets, chats and email threads sit next to your Gong calls. Customer feedback that never makes it onto a sales call still counts.
- CRM and product tracker in the same place. The assistant can tie a theme to the open pipeline that mentions it and to the tracker issue that addresses it. Customer evidence connects to code, and shipped work can be checked against what customers actually asked for.
- Evidence you can check. Every insight links to its source conversation and the exact point in the transcript. Every answer can be traced back to a customer's own words.
- Organise as you go. Labels created through MCP become filters in other tools, so an assistant's analysis builds up instead of starting from scratch each session.
The short version: Gong's MCP gives you Gong's AI-written answers, one record at a time. Four/Four's is a voice-of-customer engine your AI tools can query, and it can run on those same Gong calls.