Facebook Messenger Bot with GPT-4 for Text, Image & Voice Processing
How it Works This workflow lets you build a Messenger AI Agent capable of understanding text, images, and voice notes, and replying intelligently in real time. It starts by receiving messages from a Facebook Page via a Webhook, detects the message type (text, image, or audio), and routes it through the right branch. Each input is then prepared as a prompt and sent to an AI Agent that can respond using text generation, perform quick calculations, or fetch information from Wikipedia. Finally, the answer is formatted and sent back to Messenger via the Graph API, creating a smooth, fully automated chat experience. Set Up Steps Connect credentials Add your OpenAI API key and Facebook Page Access Token in n8n credentials. Plug the webhook Copy the Messenger webhook URL from your workflow and paste it into your Facebook Page Developer settings (Webhook → Messages → Subscribe). Customize the agent Edit the System Message of the AI Agent to define tone, temperature, and purpose (e.g. “customer support”, “math assistant”). Enable memory & tools Turn on Simple Memory to keep conversation context and activate tools like Calculator or Wikipedia. Test & deploy Switch to production mode, test text, image, and voice messages directly from Messenger. Benefits 💬 Multi-modal Understanding — Handles text, images, and audio messages seamlessly. ⚙️ Full Automation — End-to-end workflow from Messenger to AI and back. 🧠 Smart Replies — Uses OpenAI + Wikipedia + Calculator for context-aware answers. 🚀 No-Code Setup — Build your first Messenger AI in less than 30 minutes. 🔗 Extensible — Easily connect more tools or APIs like Airtable, Google Sheets, or Notion.
By Stéphane BordasWeb research assistant: automated search & scraping with Gemini AI and spreadsheet reports
⚠️ IMPORTANT: This template requires self-hosted n8n hosting due to the use of community nodes (MCP tools). It will not work on n8n Cloud. Make sure you have access to a self-hosted n8n instance before using this template. Overview This workflow automation allows a Google Gemini-powered AI Agent to orchestrate multi-source web intelligence using MCP (Model Context Protocol) tools such as Firecrawl, Brave Search, and Apify. The system allows users to interact with the agent in natural language, which then leverages various external data collection tools, processes the results, and automatically organizes them into structured spreadsheets. With built-in memory, flexible tool execution, and conversational capabilities, this workflow acts as a multi-agent research assistant, capable of retrieving, synthesizing, and delivering actionable insights in real time. How the system works AI Agent + MCP Pipeline User Interaction A chat message is received and forwarded to the AI Agent. AI Orchestration The agent, powered by Google Gemini, decides which MCP tools to invoke based on the query. Firecrawl-MCP: Recursive web crawling and content extraction. Brave-MCP: Real-time web search with structured results. Apify-MCP: Automation of web scraping tasks with scalable execution. Memory Management A memory module stores context across conversations, ensuring multi-turn reasoning and task continuity. Spreadsheet automation Results are structured in a new, automatically created Google Spreadsheet, enriched with formatting and additional metadata. Data processing The workflow generates the spreadsheet content, updates the sheet, and improves results via HTTP requests and field edits. Delivery of results Users receive a structured and contextualized dataset ready for review, analysis, or integration into other systems. Configuration instructions Estimated setup time: 45 minutes Prerequisites Self-hosted n8n instance (v0.200.0 or higher recommended) Google Gemini API key MCP-compatible nodes (Firecrawl, Brave, Apify) configured Google Sheets credentials for spreadsheet automation Detailed configuration steps Step 1: Configuring the AI Agent AI Agent node: Select Google Gemini as the LLM model Configure your Google Gemini API key in the n8n credentials Set the system prompt to guide the agent's behavior Connect the Simple Memory node to enable context tracking Step 2: Integrating MCP Tools Firecrawl-MCP Configuration: Install the @n8n/n8n-nodes-firecrawl-mcp package Configure your Firecrawl API key Set crawling parameters (depth, CSS selectors) Brave-MCP configuration: Install the @n8n/n8n-nodes-brave-mcp package Add your Brave Search API key Configure search filters (region, language, SafeSearch) Apify-MCP configuration: Install the @n8n/n8n-nodes-apify-mcp package Configure your Apify credentials Select the appropriate actors for your use cases Step 3: Spreadsheet automation “Create Spreadsheet” node: Configure Google Sheets authentication (OAuth2 or Service Account) Set the file name with dynamic timestamps Specify the destination folder in Google Drive “Generate Spreadsheet Content” node: Transform the agent's outputs into tabular format Define the columns: URL, Title, Description, Source, Timestamp Configure data formatting (dates, links, metadata) “Update Spreadsheet” node: Insert the data into the created sheet Apply automatic formatting (headers, colors, column widths) Add summary formulas if necessary Step 4: Post-processing and delivery “Data Enrichment Request” node (formerly “HTTP Request1”): Configure optional API calls to enrich the data Add additional metadata (geolocation, sentiment, categorization) Manage errors and timeouts “Edit Fields” node: Refine the final dataset (metadata, tags, filters) Clean and normalize the data Prepare the final response for the user Structure of generated Google Sheets Default columns | Column | Description | Type | |---------|-------------|------| | URL | Data source URL | Hyperlink | | Title | Page/resource title | Text | | Description | Description or content excerpt | Long text | | Source | MCP tool used (Brave/Firecrawl/Apify) | Text | | Timestamp | Date/time of collection | Date/Time | | Metadata | Additional data (JSON) | Text | Automatic formatting Headings: Bold font, colored background URLs: Formatted as clickable links Dates: Standardized ISO 8601 format Columns: Width automatically adjusted to content Use cases Business and enterprise Competitive analysis combining search, crawling, and structured scraping Market trend research with multi-source aggregation Automated reporting pipelines for business intelligence Research and academia Literature discovery across multiple sources Data collection for research projects Automated bibliographic extraction from online sources Engineering and development Discovery of APIs and documentation Aggregation of product information from multiple platforms Scalable structured scraping for datasets Personal productivity Automated creation of newsletters or knowledge hubs Personal research assistant compiling spreadsheets from various online data Key features Multi-source intelligence Firecrawl for deep crawling Brave for real-time search Apify for structured web scraping AI-driven orchestration Google Gemini for reasoning and tool selection Memory for multi-turn interactions Context-based adaptive workflows Structured data output Automatic spreadsheet creation Data enrichment and formatting Ready-to-use datasets for reporting Performance and scalability Handles multiple simultaneous tool calls Scalable web data extraction Real-time aggregation from multiple MCPs Security and privacy Secure authentication based on API keys Data managed in Google Sheets / n8n Configurable retention and deletion policies Technical architecture Workflow User query → AI agent (Gemini) → MCP tools (Firecrawl / Brave / Apify) → Aggregated results → Spreadsheet creation → Data processing → Results delivery Supported data types Text and metadata from crawled web pages Search results from Brave queries Structured data from Apify scrapers Tabular reports via Google Sheets Integration options Chat interfaces Web widget for conversational queries Slack/Teams chatbot integration REST API access points Data sources Websites (via Firecrawl/Apify) Search engines (via Brave) APIs (via HTTP Request enrichment) Performance specifications Query response: < 5 seconds (search tasks) Crawl capacity: Thousands of pages per run Spreadsheet automation: Real-time creation and updates Accuracy: > 90% when using combined sources Advanced configuration options Customization Set custom prompts for the AI Agent Adjust the spreadsheet schema for reporting needs Configure retries for failed tool runs Analytics and monitoring Track tool usage and costs Monitor crawl and search success rates Log queries and outputs for auditing Troubleshooting and support Timeouts: Manually re-run failed MCP executions Data gaps: Validate Firecrawl/Apify selectors Spreadsheet errors: Check Google Sheets API quotas
By franck fambou