Auto Reply to Facebook Comments with OpenRouter + n8n

Auto-reply to Facebook comments with an OpenRouter-powered AI Agent—powered by your own knowledge base

Turn Facebook Page conversations into an always-on support channel. This n8n workflow automatically replies to incoming Facebook comments using an OpenRouter AI Agent, backed by Postgres chat memory and a Supabase vector knowledge base that stays synced with PDFs in Google Drive.

What this workflow does

• Verifies Facebook webhooks: Handles Meta webhook verification by returning the required hub.challenge value when needed.
• Processes only real comments: Extracts incoming comment details, ignores non-comment events, and skips comments authored by the Facebook Page itself to prevent reply loops.
• Fetches context from the original post: Uses the Facebook Graph API to retrieve the original Facebook post message that the comment belongs to.
• Generates a concise reply: Uses an AI Agent powered by OpenRouter, using Postgres chat memory (per post and user) and a Supabase vector-store retrieval tool to draft an on-topic response.
• Posts the reply back to Facebook: Sends the AI-generated text as a comment reply to the specific user comment via the Facebook Graph API.
• Keeps knowledge up to date: Monitors Google Drive folders for PDF creations, updates, and “trash” events—extracting text, generating OpenAI embeddings, and inserting/deleting matching vectors in Supabase so the AI stays current.

Use cases

• Automated, consistent responses to FAQ-style Facebook comments for SaaS products.
• Support teams using a PDF knowledge base (policies, onboarding docs, product docs) without manual copy/paste.
• Page admins reducing response time while keeping replies grounded in your latest documentation.

Technical details

• Integrations: Meta Facebook (webhooks + Facebook Graph API), OpenRouter, Postgres, Supabase, Google Drive, and PDFs.
• n8n nodes/tools noted in workflow: webhook, if, set, no op, supabase, and sticky note.
• Memory + retrieval: Postgres chat memory plus Supabase vector-store retrieval for knowledge-grounded replies.