01
Turn on Cloud MCP in Kaption
In WhatsApp Web with the Kaption extension (Chrome or Edge), open AI Assistants (MCP) › Cloud Bridge › Enable. In the Kaption desktop app, switch on Cloud MCP in Settings. Keep Kaption open while ChatGPT works.
MCP · ChatGPT
Read WhatsApp voice notes with ChatGPT
Kaption turns voice notes into text in WhatsApp Web and the desktop app. Connect it to ChatGPT and ask what someone said in their voice notes, find one by topic, or get the key points of a long one.
MCP server address
https://mcp.kaptionai.com/mcpShort answer
Can ChatGPT transcribe and summarize my WhatsApp voice notes?
Kaption does the transcription and ChatGPT does the reading. Transcribe voice notes with Kaption in WhatsApp Web or the desktop app (unlimited in your first 60 days; after that, Free continues with limits on voice notes), connect Kaption’s MCP server to ChatGPT, and ask “what did Leo say in his voice notes today?”. ChatGPT reads the transcripts through Kaption on your computer and summarizes them, within your plan’s monthly credits.
What do I need to do this in ChatGPT?
Kaption open on your computer (the extension in WhatsApp Web, in Chrome or Edge, or the Kaption desktop app) with Cloud MCP on, a Kaption account (every account starts with 60 days of Business, unlimited; after that, a plan with credits), and ChatGPT. Setup takes three steps:
- 02
Create the app in ChatGPT
Turn on Developer mode in Settings › Apps › Advanced settings (Connectors on some plans), click Create, name it WhatsApp, paste https://mcp.kaptionai.com/mcp and choose OAuth.
- 03
Sign in and pick it in a chat
Enter your phone number and the 6-digit code Kaption sends you on WhatsApp. Then pick the WhatsApp app from the tools menu in a ChatGPT conversation.
What should I ask ChatGPT?
Copy a prompt, paste it into ChatGPT and change the names to your own chats and contacts.
What did Leo say in his voice notes today? Summarize each one in a line.
Find the voice note where my accountant talked about the tax deadline and tell me the date she gave.
Summarize the voice notes in the Team group since yesterday and list any tasks for me.
Search my voice-note transcripts for “contract” and tell me who mentioned it and when.
Turn my contractor’s voice notes from this week into a checklist of what he said he would do, and by when.
What can ChatGPT do here, and what can’t it?
It can
- Read the transcript of any voice note Kaption has transcribed, sent or received, in the language it was recorded in.
- Search across transcripts by word, person or chat.
- Summarize one long voice note or a day of them, and pull out dates, prices and tasks.
- Answer in your language, even when the voice note was recorded in another one.
- Work wherever your ChatGPT plan and workspace allow custom MCP apps. If you don’t see Developer mode, ask your workspace admin or check OpenAI’s help center.
It can’t
- Listen to audio. ChatGPT works from Kaption’s transcripts, so a voice note with no transcript has nothing to read: transcribe it in Kaption first.
- Fix what the transcript got wrong. Background noise, crosstalk and unusual names can come out wrong; play the audio when a number or name really matters.
- Read voice notes in locked chats.
- Reach your WhatsApp while Kaption is closed.
What leaves my computer?
Transcription happens in Kaption: only the audio you choose to transcribe leaves your computer, without your name, number or contacts, and it is discarded once the text comes back. When ChatGPT reads a voice note, Kaption sends it the transcript.
Kaption’s relay at mcp.kaptionai.com checks your sign-in and passes each request to Kaption on your computer, which runs it against your WhatsApp and sends back only the answer; the relay does not keep your chats. What comes back, such as message text, names and numbers, goes to ChatGPT, which handles it under its own privacy terms. Locked chats are never visible to the assistant, and you can disconnect at any time.
How many credits does my plan include?
Claude and ChatGPT are unlimited in every account’s first 60 days; after that they need a plan, counted in credits per calendar month: Lifetime local AI 50, Pro 250, and Business 500 with its 2 numbers plus 250 for each extra number, shared by all its numbers, the plan to pick when an assistant uses WhatsApp every day. Business Agent includes them with fair use. A read uses 1 credit, a send 2 and a summary 3. See plans and prices
Questions
Do voice notes need a transcript first?
Yes. ChatGPT reads text, so a voice note needs a Kaption transcript before ChatGPT can read it. Transcribe it in WhatsApp Web or the desktop app: unlimited in every account’s first 60 days; after that, Free continues with limits on voice notes.
Which languages?
Kaption transcribes dozens of languages, each note in the language it was recorded in, and ChatGPT can answer you in yours.
Can it read voice notes I sent?
Yes. Kaption transcribes any voice note in a chat, sent or received, and ChatGPT can read either.
Which ChatGPT plan do I need?
One that lets you add a custom MCP app with developer mode. OpenAI changes which plans and workspaces allow it, so if the option is missing, check OpenAI’s help center or ask your workspace admin.
Connect WhatsApp to ChatGPT
Turn on Cloud MCP in Kaption, add the address to ChatGPT and sign in with a code on WhatsApp.
Set up WhatsApp in ChatGPT