Gong MCP Server

An MCP (Model Context Protocol) server that provides access to your Gong.io data. Query calls, transcripts, users, keyword trackers, and more directly from Claude or any MCP-compatible client.

Tools Quick Reference

Response Size & Context Limits

search_calls can return a lot of data. Under the hood it:

• Auto-paginates up to ~5000 calls (50 API pages) per query
• Applies client-side filters (participant, customer, tracker, duration, etc.) after pagination
• Returns a rich per-call format (metadata + summary + topics + participants) by default

Guardrails built in:

• If the formatted output would exceed MAX_MCP_OUTPUT_LENGTH (default 50000 chars, configurable via env var), the tool automatically falls back to a compact table with a warning. You still get every call ID and title — drill in with get_call_summary on specific ones.
• include: ["outline"] is expensive (~80KB per call). Avoid it in multi-call searches.
• Tracker data is filtered to only show trackers matching your trackers filter (or non-zero trackers if no filter) — no more walls of (0x) noise.

If your query hits the output cap, narrow it:

1. Tighten fromDateTime / toDateTime
2. Add scope: "External" or scope: "Internal"
3. Add minDuration: 600 to skip short no-shows
4. Add customerName or trackers filter
5. Drop heavy include options like outline

Prerequisites

• Node.js 18+ or Docker
• Gong API credentials (Access Key and Secret)

1. Log into Gong as an admin
2. Go to Company Settings → Ecosystem → API
3. Click Create API Key
4. Save both the Access Key and Secret (the secret is only shown once)

Option 1: npx (no install required)

npx gongio-mcp

Option 2: Global npm install

npm install -g gongio-mcp
gongio-mcp

Option 3: From source

git clone https://github.com/JustinBeckwith/gongio-mcp.git
cd gongio-mcp
npm install
npm run build
node dist/index.js

Option 4: Docker (build locally)

git clone https://github.com/JustinBeckwith/gongio-mcp.git
cd gongio-mcp
docker build -t gongio-mcp .
docker run --rm -i \
  -e GONG_ACCESS_KEY=your-access-key \
  -e GONG_ACCESS_KEY_SECRET=your-secret-key \
  gongio-mcp

Set your Gong credentials as environment variables:

export GONG_ACCESS_KEY="your-access-key"
export GONG_ACCESS_KEY_SECRET="your-secret-key"

Or pass them inline:

GONG_ACCESS_KEY=your-key GONG_ACCESS_KEY_SECRET=your-secret npx gongio-mcp

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

Using npx:

{
  "mcpServers": {
    "gong": {
      "command": "npx",
      "args": ["gongio-mcp"],
      "env": {
        "GONG_ACCESS_KEY": "your-access-key",
        "GONG_ACCESS_KEY_SECRET": "your-secret-key"
      }
    }
  }
}

Using Docker:

{
  "mcpServers": {
    "gong": {
      "command": "docker",
      "args": ["run", "--rm", "-i",
               "-e", "GONG_ACCESS_KEY",
               "-e", "GONG_ACCESS_KEY_SECRET",
               "gongio-mcp"],
      "env": {
        "GONG_ACCESS_KEY": "your-access-key",
        "GONG_ACCESS_KEY_SECRET": "your-secret-key"
      }
    }
  }
}

Claude Code

Using npx:

claude mcp add gong -e GONG_ACCESS_KEY=your-key -e GONG_ACCESS_KEY_SECRET=your-secret -- npx gongio-mcp

Using Docker (after docker build -t gongio-mcp .):

claude mcp add gong -e GONG_ACCESS_KEY=your-key -e GONG_ACCESS_KEY_SECRET=your-secret -- docker run --rm -i -e GONG_ACCESS_KEY -e GONG_ACCESS_KEY_SECRET gongio-mcp

Available Tools

<a name="list_calls"></a>

List calls with optional date range and workspace filters. Returns minimal call metadata (ID, title, date, duration).

Parameters:

<a name="get_call"></a>

Get the URL, timing, direction, scope, system, and other metadata for one call. Faster than get_call_summary when you only need call metadata.

Parameters:

<a name="get_call_summary"></a>

Get an AI-generated summary including brief overview, key points, topics, action items, and detailed outline. This is the recommended way to understand a call — use get_call_transcript only if you need exact quotes.

Parameters:

<a name="get_call_transcript"></a>

Get the raw transcript with speaker attribution. Transcripts are paginated (default 10KB) to prevent context overflow — use maxLength and offset to navigate.

Parameters:

<a name="search_calls"></a>

Search calls with advanced filters including participant lookup, customer name search, and rich content selection. Automatically paginates through all results and returns participant info, brief summary, and topics by default.

Parameters:

Date & workspace

Host / participant

Content & customer

Metadata

Response shape

include options:

Defaults (always returned): participants, brief summary, and topics — ~3KB per call.

Filter behavior notes:

• Filters combine with AND logic. primaryUserIds + participantUserIds compose: primary narrows server-side, participant post-filters.
• customerName matches any of: CRM account Name field, external participant email domain, or call title (case-insensitive substring).
• trackers names are workspace-specific — call get_trackers first to see what's configured. Match is case-insensitive substring (so "competitor" matches both "Competitors" and "Competitor Mentions").
• When trackers filter is set, the relevant tracker content is auto-included and the output shows only the trackers you asked about. Without a filter, only non-zero trackers are shown.
• An empty result returns "No calls found" rather than an error.

See Response Size & Context Limits for how large-result fallback works.

<a name="search_calls_by_account"></a>

Find calls involving a specific account or company by matching the email domains of external participants. The Gong API does not natively support filtering by account name (a known gap) — this tool fetches calls in the date range and post-filters on parties[].emailAddress. Auto-paginates the underlying /v2/calls/extensive endpoint up to maxCalls.

Use this when:

• A prospect has multiple email domains (acme.com, acme.io, regional TLDs) and you need them all
• You need to join external enrichment data (e.g., "all prospects on Klaviyo" from BuiltWith / Clearbit / a vendor-stack graph) — resolve to a domain list upstream and pass it here
• Domain-based matching is more reliable than CRM Account names that drift across systems

Parameters:

Cost note: This is fetch-then-filter. A 90-day window with no other narrowing typically pages through 1–5 API calls. Combine with primaryUserIds to bound cost on long ranges.

<a name="search_calls_by_opportunity"></a>

Find calls linked to specific CRM Opportunities by ID or name substring. Requires Gong-CRM integration (Salesforce / HubSpot) — calls without CRM linkage will not match.

Use this when:

• You want every call on a specific deal — opportunityIds: ["006xxxxx"] is the most precise option
• Opportunity names are descriptive (e.g. "Acme Q4 Renewal") and you want fuzzy matching across renamed/duplicated opportunities

Parameters:

Cost note: Uses the same fetch-then-filter pattern as search_calls_by_account. CRM context lookup adds no extra API calls — it rides on the same /v2/calls/extensive request with context: "Extended".

<a name="search_transcripts"></a>

Free-text keyword search across call transcript sentences within a bounded date range. Two-phase: (1) /v2/calls/extensive narrows the call set by date + optional primaryUserIds / domains, (2) /v2/calls/transcript fetches transcripts for the narrowed set and returns sentence-level matches with speaker attribution and timestamps.

Prefer Gong Trackers for recurring terms. For competitor names, ESP/tech terms (Klaviyo, Braze, Iterable, Postscript, Attentive, Sendgrid, Customer.io, etc.), and other terms you'll search for repeatedly — set them up as Gong Trackers in the UI (one-time, ~30 minutes for ~20 terms). Then use get_trackers + search_calls + get_call_summary instead. Trackers are server-side, the cost is dramatically lower, and they surface counts and timestamps natively. Use search_transcripts for ad-hoc one-offs.

Cost guard: Date ranges greater than 30 days require additional narrowing via primaryUserIds or domains. A 6-month unbounded scan would burn API quota and is rejected at the schema level.

Parameters:

Returns: Sentence-level matches grouped by call, including speaker name and affiliation (when available), keyword matched, timestamp (mm:ss), and the sentence snippet.

<a name="get_trackers"></a>

List all keyword tracker definitions including tracked phrases, affiliation (whose speech is tracked), and filter queries. Explains tracker hits visible in get_call_summary output.

Parameters:

<a name="list_workspaces"></a>

List all Gong workspaces with their IDs and names. Use these IDs as filters in list_calls, search_calls, get_trackers, and other tools. Most companies have 1–3 workspaces (e.g., by region or product line).

Parameters: None

<a name="get_user"></a>

Get a specific user's profile including name, email, title, phone, and settings. Useful for resolving user IDs returned from call data.

Parameters:

<a name="search_users"></a>

Search and filter users by IDs or creation date. More flexible than list_users for resolving specific user IDs from call data in bulk.

Parameters:

<a name="list_users"></a>

List all Gong users in your workspace. Returns name, email, and title for each user.

Parameters:

<a name="list_library_folders"></a>

List all public Gong call library folders for a workspace. Returns folder IDs and names used with get_library_folder_calls. Private and archived folders are not returned.

Parameters:

<a name="get_library_folder_calls"></a>

Get all calls saved in a specific Gong library folder. Returns call IDs, titles, curator notes, and snippet timing for clips. Call IDs can be passed directly to get_call_summary or get_call_transcript.

Parameters:

Available Resources

<a name="gong-users"></a>

Returns a markdown-formatted list of all users in your Gong workspace. Useful for resolving user IDs found in call data.

Parameters: None

Example Prompts

Once connected to Claude, you can ask:

• "List my Gong calls from last week"
• "Get the details for call 123456789"
• "Show me a summary of call 123456789"
• "Get the transcript for call 789012"
• "What workspaces do we have in Gong?"
• "What keywords is Gong tracking for competitors?"
• "What call library folders do we have in Gong?"
• "Show me the calls in the 'Best Discovery Calls' library folder"
• "Who are all the users in our Gong workspace?"
• "Search for calls hosted by Justin (user ID 232255198215877499) in July 2025"
• "Look up these user IDs: 111, 222, 333"
• "Show me all calls from the past 60 days with anyone at acme.com or acme.io"
• "Find every call attached to opportunity 006xxxxx"
• "Find Q3 calls where prospects mentioned Klaviyo or Braze, narrowed to John's calls"

Contributing

Interested in contributing? Check out CONTRIBUTING.md for development setup, testing instructions, and guidelines.

The repository includes gong-openapi.json — a local copy of the Gong API OpenAPI spec. It's useful as a reference when adding new tools: use it to look up endpoint paths, parameter names, and response shapes without leaving your editor. The latest spec can be downloaded from the Gong API documentation.