Multi-Modal AI Expense Tracking Assistant using Telegram and Google Sheets - n8n Workflow

• Users who want effortless financial tracking and automation.


• Technical users seeking advanced n8n templates featuring complex AI agents and tool use.


• Individuals needing multi-currency expense logging capabilities directly from a chat app.


• Automation specialists exploring multi-modal input handling (voice, image, text) within an n8n workflow.

This sophisticated n8n workflow addresses the pain points of manual expense tracking by establishing a fully automated, conversational financial assistant accessible via Telegram. This robust n8n workflow leverages multiple specialized AI models, including Google Gemini for receipt OCR and Azure OpenAI (GPT-4) for classification, analysis, and categorization, enabling seamless logging from voice notes, images of receipts, and simple text commands. The core of this system is a powerful AI Agent n8n node that utilizes a suite of tools accessible via the Multi-Channel Proxy (MCP) pattern. These tools include Google Sheets n8n node operations for secure data storage and an Exchange Rate n8n node for real-time currency conversion, ensuring highly accurate record-keeping. By using this n8n template, users gain instant access to spending summaries, balance checks, and detailed financial reports, making it one of the most advanced financial n8n templates available for personalized automation.

The entire process is initiated by a single n8n trigger and involves intelligent routing and specialized AI models to manage complex financial data.

1. Start and Input Trigger: The n8n workflow begins with the Telegram Trigger n8n node, which activates upon receiving any new message, voice note, or image from the user.


2. Multi-Modal Routing: A Switch n8n node (Route Input by Message Type) analyzes the payload. Voice notes are sent to the ElevenLabs n8n node for transcription. Receipt images are downloaded and processed by the Google Gemini n8n node to perform OCR and extract the amount and description.


3. Intent Classification: All inputs are consolidated into text and fed into the Classify Message Intent n8n node, which uses a GPT-4 model to quickly classify the user's intent as either expenses (logging a purchase) or other (asking a query, like for a balance).


4. Expense Formatting: If the input is classified as expenses, the Parse and Split Multiple Expenses n8n node (another GPT-4 model) processes the text. This crucial n8n workflow step cleans up the entry and splits complex, multi-item messages (e.g., "coffee 5 and gas 15") into separate, structured inputs for accurate logging.


5. The AI Agent Core: The formatted request proceeds to the main Agent n8n node. This agent, powered by GPT-4 and extensive system instructions, determines the necessary steps (e.g., currency conversion, calculation, or data lookup).


6. Tool Execution via MCP: The Agent communicates with specialized tools (Google Sheets, Calculator, Exchange Rate) using the MCP Client/Server n8n node pattern. For instance, to log an expense, it uses the ExchangeRate tool, then the Calculator tool, and finally the AddExpenses_Tool n8n node.


7. Final Response: Once the action is complete—whether logging an expense or generating a financial report—the Agent crafts a user-friendly response, which is then delivered back to the user through the Send Message Telegram n8n node.

To deploy this comprehensive n8n workflow, you must complete the following steps:

1. Import the n8n Workflow: Copy the provided JSON data and paste it directly into your n8n instance using the Import from JSON feature.


2. Telegram Setup: Configure the Telegram Trigger n8n node and the Send Message n8n node with your Telegram Bot Token credentials.


3. Google Sheets Setup: Set up a Google Sheets credential and link it to the four Google Sheets n8n node tools (TotalSpentTool, BalanceTool, ExpensesTool, AddExpensesTool, ReadRowsTool). Ensure your spreadsheet has two tabs named Expenses and Balance&TotalSpent with the required header columns as described in the workflow notes.


4. AI Credentials: Provide credentials for Azure OpenAI (or standard OpenAI if adapting the LLM n8n node), ensuring access to GPT-4 for the Agent, Intent Classifier, and Parser nodes. Also, configure the Google Gemini n8n node credentials for receipt OCR.


5. ElevenLabs & Exchange Rate API: Configure the Convert Voice to Text (ElevenLabs) n8n node and update the URL in the ExchangeRate HTTP Request n8n node with your required API key.


6. Activate: After securing all credentials and ensuring the Google Sheets n8n node configurations point to the correct sheet IDs and names, activate the Telegram Trigger n8n node to start the automation.

Telegram Trigger (n8n trigger): Starts the n8n workflow upon receiving a message, voice note, or image. Essential for the conversational interface.
Google Gemini (n8n node): Used in the Extract Receipt Data step. It performs multi-modal analysis (OCR) on uploaded images to extract expense amounts and descriptions.
Convert Voice to Text (ElevenLabs HTTP Request n8n node): Transcribes audio messages (voice notes) into text, enabling voice-based expense logging.
Intent Classification Model (Azure OpenAI/GPT-4 n8n node): A specialized language model that routes the n8n workflow by quickly classifying user input as either a new expense or a data query.
AI Agent (GPT-4 Agent n8n node): The core intelligence of the n8n workflow. It uses a detailed system prompt and accesses multiple tools (via MCP) to execute complex financial tasks, ensuring accurate categorization and logging.
MCP Server Trigger & MCP Client (n8n node): These nodes implement the Multi-Channel Proxy pattern, allowing the central AI Agent to securely call specialized tools like the Calculator and Google Sheets functions without direct connections, simplifying the Agent’s setup.
AddExpensesTool (Google Sheets n8n node): The primary data logger. It appends the categorized and structured expense data into the designated Google Sheet, completing the logging phase of the n8n workflow.
Exchange_Rate (HTTP Request n8n node): Fetches real-time currency conversion rates, allowing the AI to automatically convert non-USD expenses to USD before logging.