Gong

What it is

Gong records, transcribes and analyzes customer-facing conversations — calls, video meetings, email — and turns them into structured signals: risk flags on deals, coaching moments for reps, market intelligence for execs. Around that core it sells four modules: Gong Engage (sales engagement), Gong Forecast (pipeline and revenue prediction), Gong Enable (enablement and scorecards) and the Gong Revenue Graph, the data layer the rest reads from.

Two releases in 2026 changed what you are buying. Mission Andromeda on 25 February 2026 shipped Gong Enable, the Gong Assistant and — the change that matters most for a 2026 stack — “built-in client and server support for MCP.” Mission Big Dipper on 24 June 2026 added the Gong Revenue Harness, an execution layer that governs and orchestrates agents, plus Custom Agents, an AI Builder inside Gong Assistant and an AI Coach in Enable. Gong now markets itself as a Revenue AI OS rather than a conversation-intelligence tool.

The commercial position behind that is real, not a repositioning story. Gong reported ARR past $500M growing over 55% year over year on 12 May 2026, its tenth consecutive quarter of accelerating growth, across more than 5,000 customers including Google, Cisco, ADP, Docusign, Thomson Reuters, Canva and Anthropic.

Why it shows up in RevOps stacks

Conversation intelligence is still the moat. Years of training on B2B sales conversations put Gong’s transcription, topic extraction and risk-flagging ahead of generic LLM transcription, and the gap shows up most on the boring parts: speaker separation on a six-person call, and catching a competitor mention in the ninth minute of a 50-minute recording.

The MCP server closed the one gap Gong had. Until this cycle, the argument for Chorus or Attention over Gong was agent access: they exposed a server and Gong didn’t. That is over. Gong ships a hosted MCP server and an MCP client, and the server exposes three read-only tools — ask_account, ask_deal and generate_brief — that answer natural-language questions about one account or one deal and return a structured brief covering themes, stakeholders, risks and next steps. A tech admin enables it under Company Settings → Ecosystem → API → Integrations → Create Integration → MCP, and per Gong’s docs it is available on any Gong plan, gated on the seat you are assigned rather than on a paid upgrade. Connecting it to Claude needs a Pro, Max, Team or Enterprise plan on the Claude side.

Read the tool list carefully before you budget against it. Three read-only tools is a briefing surface, not a workspace — you can ask Gong about a deal from inside an assistant, but you cannot make it change anything. Attention’s server exposes 68 tools across read and write scopes. If the requirement is “the assistant updates the record,” Gong’s server does not do it.

Forecast triangulation. Gong Forecast pulls CRM, calendar, email and conversation signals into one deal-health model. Most RevOps leaders run it as a check against the rep-submitted number in the CRM, not as a replacement for it.

Pricing reality

Quote-only, and unchanged on that point: gong.io/pricing published no figures and no tier names when checked on 2 September 2026. The page states only that “licenses are priced per user,” that “there is a platform fee based on the number of users supported,” and that integrations are included at no extra charge. The form segments you by team size — 1-50, 51-1,000, 1,001-9,999, 10,000+ — before anyone quotes you.

Every number below is triangulated from procurement data and marked as such. Negotiate against them; do not quote them back as list price.

• Total contract value. Vendr’s marketplace dataset puts the median Gong contract at roughly $55,000/year across 1,131 purchases, in an $11,230-$204,022 range, at an average negotiated discount near 14%.
• Per seat. Approximately $1,400-$2,000 per seat/year on a standard-tier, 25-75 seat, one-year deployment; approximately $1,800-$2,600 at the higher tier on 50-150 seat contracts. The wide band is real — the same seat count quotes differently depending on module mix.
• The platform fee is the line that surprises people. It is a separate mandatory annual charge on top of licenses, reported in the $5,000-$50,000 range depending on headcount. On a 10-seat deal it can add $500/user/year before a single license, which is why Gong economics invert on small teams.
• Modules are priced on top. Gong Engage and Gong Forecast typically add 20-40% of base platform cost per seat, and bundling them at initial purchase prices materially better than adding them mid-contract.
• Seat minimums and terms. Annual only, no monthly option, with a practical floor around 15-25 seats. Multi-year commitments and quarter-end timing are where the 15-30% discounts actually live.

The MCP server does not carry its own SKU. If a rep tells you agent access is an upgrade, point at Gong’s own documentation, which puts it on any plan.

Best for

A RevOps leader or CRO at a $20M+ ARR B2B company with 40 or more quota-carrying reps, where coaching is a funded program with an owner rather than a manager’s side task. The scoped case where Gong wins outright: a multi-segment sales org that needs one conversation record of truth across new business, expansion and renewal, and now wants Copilot, Agentforce or Claude to answer deal questions against that record without exporting anything.

Not for

Teams under 15 reps, and this is the anti-ICP worth stating plainly. The platform fee is close to fixed while your license count is small, so cost per rep runs at its worst exactly where budgets are tightest. Under 15 reps, Fathom, Granola or Fireflies at $0-30/user/month plus Claude-side analysis on the transcripts gets most of the decision value for roughly a tenth of the money. It is also the wrong buy when nobody owns the output — Gong surfaces more signal than an org without a dedicated enablement or RevOps owner will ever act on, and unread dashboards renew at full price.

Versus the alternatives

• Chorus (ZoomInfo) — pick it when ZoomInfo is already your contact-data layer, because Chorus is an add-on line on that platform contract rather than a second platform. Vendr’s median ZoomInfo contract is about $33,500/year including the data, against Gong’s $55,000 median for conversation intelligence alone. Pick Gong when coaching workflow and deal inspection are the deliverable; Chorus trails on both.
• Clari — pick it when the board conversation is forecast accuracy and the buyer is the CRO, especially since Salesloft merged into Clari in December 2025 and the engagement layer now sits inside the same contract. Clari Core triangulates near $1,200-$1,500 per user/year with a ~$76,000 median contract. Gong is the stronger call-level product; Clari is the stronger number-level one. Above roughly 150 reps, teams commonly run both.
• Attention — the fastest-growing entrant in the segment ($30M Series B on 24 June 2026, ARR up 4x). Pick it when the deliverable is the follow-up email sent and the CRM field written, not the dashboard that reports they weren’t. Its MCP server exposes 68 tools with a write scope against Gong’s 3 read-only ones. Pick Gong when scale, procurement comfort and coaching depth across 100+ reps outweigh write access.

If none of these fit, the honest answer is usually that the problem is not visibility. Buy a $20/seat notetaker, spend the difference on a single enablement hire, and revisit in two quarters.

Watch-outs

• The platform fee makes small deployments quietly terrible value. It scales with supported users but does not fall to zero as your seat count does. Guard: get the platform fee quoted as a separate line before you discuss per-seat price, and compute cost-per-rep including it. If that number exceeds roughly $2,500/rep/year, you are below the size band where this product pays back.
• The three read-only MCP tools are a briefing surface, not automation. Buying Gong expecting an assistant to update deals ends in a build project against the REST API. Guard: write the three tool names into the trial plan and test the exact question you intend to ask in production. If the requirement includes a write, scope that against the API — or against Attention — before signing.
• The MCP server inherits the authorizing user’s visibility. Personal Access means Claude sees precisely what that person sees, which is a feature until a sales leader with org-wide visibility authorizes the connector for convenience. Guard: authorize from a purpose-built account with the narrowest seat that satisfies the use case, and log who created the integration under Company Settings → Ecosystem → API. Review it on the same cadence as CRM profile changes.
• Total cost of ownership is analyst time, not just license spend. Gong surfaces more than a part-time owner can process, and the coaching ROI everyone models assumes someone runs the program. Guard: name the owner and the weekly hours in the business case. If nobody can commit 4-6 hours a week, cut the seat count instead of the price.
• Recording is a compliance project in regulated industries and multi-jurisdiction teams. Two-party consent rules and works-council review in DACH routinely add weeks. Guard: start legal and works-council review the week you start the trial, not the week before go-live, and pilot in a single jurisdiction until the sign-off lands.

Related: Gong vs Chorus and Gong vs Clari work the head-to-heads, alternatives to Gong covers the wider replacement set, and the conversation intelligence stack prices the layers around it.