Build a Voice‑AI Assistant with a Real‑Time ‘MCP Brain’ — Telegram + n8n + OpenAI + Google Docs

Turn Telegram into your very own voice‑powered AI side‑kick—no coding required!
In this in‑depth beginner tutorial (≈ 60 min) we wire together n8n, OpenAI, and a tool‑packing MCP “brain” so your bot can:

listen to voice notes

transcribe them with Whisper

fetch live content from Google Docs, Sheets, Gmail, Airtable, even a calculator

craft an answer with GPT‑4

and speak it back via Telegram audio

Whether you’re automating customer support, daily reports, or just geeking out—this is an end‑to‑end guide you can clone, tweak, and deploy today.

00:00 Preview Demo – ask & get a spoken answer
01:20 What We’ll Build (high‑level fly‑over)
04:10 Key Concepts Explained – webhooks, MCP, AI Agent
09:25 Creating the Telegram Bot with @BotFather
12:10 Telegram Trigger Node – catching voice messages in n8n
15:05 HTTP 1 – using file_id to fetch download path
18:00 HTTP 2 – downloading the actual .oga voice file
20:40 Whisper Transcription – OpenAI node setup
24:30 AI Agent Basics – why we need a “brain”
26:45 MCP Overview – client/server & tool anatomy
28:55 Building MCP Server: Google Docs Tool (getDocContent)
33:20 Building MCP Server: Google Sheets Tool (readSheetRow)
35:35 Gmail, Airtable & Calculator Tools (rapid walkthrough)
38:20 Wiring MCP Client nodes into the Agent
41:40 Crafting the System Prompt + Tool Descriptions
44:05 First Full Run – agent fetches doc + answers in text
46:15 OpenAI Text‑to‑Speech – choosing model, voice, OPUS format
49:00 Telegram Send Audio – delivering the voice reply
51:30 Error‑Handling & Testing Tips
54:15 Cost & Rate Limit Considerations (OpenAI & Google)
56:00 Deploying n8n on VPS / n8n.cloud
58:00 Next‑Step Ideas – memory, multi‑language, UI polish
59:30 Wrap‑up & Resources

Видео Build a Voice‑AI Assistant with a Real‑Time ‘MCP Brain’ — Telegram + n8n + OpenAI + Google Docs канала MN Marketing Solutions