See llms.txt for all machine-readable content.

Back to Templates

Write real estate listing descriptions with Claude and Slack

Quick overview

A new listing in your CRM or MLS feed triggers Claude to write the description, the email blast and the social posts in your voice, with a fair-housing check on every draft. You approve in one place; the assets go to Mailchimp, Buffer or the Meta API.

How it works

  1. Receives a POST webhook containing listing details and photo URLs.
  2. Uses Anthropic Claude to generate photo descriptions and stops for manual follow-up if the model flags the result as needing a person.
  3. Uses Anthropic Claude to draft an MLS-ready listing description and diverts flagged outputs for manual review.
  4. Uses Anthropic Claude to draft derivative assets (social captions, an email draft, and a short narration script) and diverts flagged outputs for manual review.
  5. Runs a compliance/brand check step across the generated content.
  6. Sends a Slack message for approval and waits for an Approve/Reject response.
  7. If approved, posts to Facebook via the Facebook Graph API and appends tracking data to Google Sheets.

Setup

  1. Configure the incoming webhook URL in your source system so it sends the listing payload and photo URLs to this workflow.
  2. Add Anthropic credentials for the Claude chat model used in the AI drafting steps.
  3. Add Slack credentials and set the approver channel name (or address) in the Config values used by the approval step.
  4. Add Facebook Graph API credentials and provide a valid Facebook Page ID/message fields used for publishing.
  5. Add Google Sheets credentials and replace the spreadsheet ID and sheet name used for tracking.

Requirements

  • n8n 1.x (tested on 2.38), an Anthropic API key for the Claude nodes, and Follow Up Boss / kvCORE, Mailchimp / Constant Contact, Buffer / Meta API (or their equivalents, the nodes are standard and easy to swap).

Customization

  • Put your voice rules and fair-housing checklist in the Config node; publish through Mailchimp, Buffer or the Meta API. Keep the guardrail: Fair-housing language check runs on every asset before review; flagged copy cannot be approved without an edit.

Additional info

One of 48 Agentmelt blueprints. The blueprint page explains the trigger, the steps, the guardrails and where the standard version stops: https://agentmelt.com/workflows/listing-description-automation/ and all 48 workflow files are on GitHub (MIT): https://github.com/zeshutmax/agentmelt-blueprints