Claude + Gong
Your calls already hold the truth about every deal. Wire Gong to Claude and the transcripts become follow-ups, briefs, risks, and coaching.
- Category
- Conversation Intelligence
- Connect via
- MCP · API
- Data
- Read (transcripts + deal data); writes go to your CRM, not Gong
- Stage
- Close & Expand
- Best for
- Turning what was said on calls into follow-ups, deal intelligence, and coaching
Gong, and why connect it to Claude
Gong records and transcribes your customer conversations and layers deal intelligence on top: who said what, which topics came up, where deals are drifting. It is the closest thing GTM has to a record of ground truth, because the calls are what actually happened, before anyone summarized them optimistically into the CRM.
Connecting Claude puts that ground truth to work. Instead of a rep half-remembering a call at the end of the day, the transcript becomes the input: the follow-up drafted from what was actually said, the CRM note that quotes the customer, the risk flagged because the champion said the words budget review, the coaching note grounded in the rep's real talk track.
Three ways to wire Claude to Gong
Three ways to wire Claude to Gong, from a chat connector to a fully scheduled pipeline.
As a connector
Add Gong as a connector in the Claude apps or wire it into Claude Code, authorized with your Gong login. Best for on-demand work: pull the last call with an account, draft the follow-up, prep the next conversation, from a chat.
The API, via Claude Code or the Agent SDK
For the scheduled version, call Gong's API from Claude Code or the Agent SDK: every morning, fetch yesterday's calls, draft the follow-ups and CRM notes, flag the risks, and leave the batch for review. This is where the compounding value lives.
Transcript export as the fallback
No API access yet? Export the transcript and hand Claude the file. It is manual, but the plays are identical, and it is a fine way to prove the value before anyone provisions credentials.
Read first, write on purposeGong is a read-mostly integration: the transcripts and deal data flow out, and the writes land in your CRM, not in Gong. So the gate to watch is the same CRM gate as everywhere, the note, the field update, the task Claude drafts from a call should pass your approval before it hits the record.
What you can do once they're connected
Follow-ups drafted from the transcript while the call is still warm, quoting what the customer actually said.
CRM notes that reflect the conversation instead of a rep's end-of-day reconstruction.
Deal risks surfaced from language: pricing pushback, a rival named, the champion hedging.
Coaching notes grounded in the rep's real talk track against your framework, not a vibe.
What to build with Gong and Claude
Each is a full playbook or skill you can ship, wired to Gong.
Post-call follow-up + CRM note
The signature play: transcript in, follow-up and CRM note out, minutes after the call.
Discovery call prep skill
Prep the next call from the last one: what was said, what to ask, what to avoid.
Mutual action plan template
Keep the MAP current from what was actually agreed on calls.
Weekly competitive intel digest
Mine the week's calls for competitor mentions and objections, digested every Friday.
Gotchas worth knowing first
- Transcripts are long and context is finite. Send the calls the task needs, not the account's whole history, or the important quote drowns in the noise.
- Transcription mangles names, numbers, and product terms. Give Claude your glossary, the product names, the competitors, the acronyms, so it reads through the typos.
- Recording consent is a legal matter that varies by region. Your Gong admin has consent settings for a reason; automating on top does not change the rules.
- Do not let auto-drafted CRM notes skip review at the start. A note that misattributes a quote reads as fact forever once it is on the record.
Claude + Gong: common questions
Does Claude integrate with Gong?
Yes. Add Gong as a connector in the Claude apps, wire it into Claude Code, or call Gong's API from the Agent SDK for scheduled pipelines. Even without credentials, an exported transcript works for the same plays.
What is the first play to run?
The post-call follow-up: transcript in, a follow-up email and CRM note out, minutes after the call. It is high value, low risk, and shows the pattern every other play builds on.
Can Claude analyze every call automatically?
Yes, on a schedule: fetch yesterday's calls each morning, draft the follow-ups and notes, flag the risks, and leave the batch for review. Claude Code or the Agent SDK plus Gong's API is the shape.
Will it get names and product terms right?
Transcripts mangle proper nouns, so give Claude a glossary of your products, competitors, and acronyms in the standing context. With it, the reasoning reads through transcription errors instead of repeating them.
Does Claude write anything back into Gong?
Generally no, Gong is the source. The outputs, follow-ups, notes, tasks, risks, land in email drafts and your CRM, which is where the approval gate belongs.