Produce illustrated, research-based newsletters from Airtable with OpenAI and Kit

Quick overview

This workflow runs every six days to pick one “Ready” newsletter topic from Airtable, research and write five cited sections with OpenAI models, generate matching images via Replicate, host them on Cloudinary, then create a draft (and optionally schedule) a broadcast in Kit.

How it works

1. Runs on a 6-day schedule, loads configuration values, and selects a single Airtable record marked Ready with a newsletter topic and no Broadcast ID.
2. Validates the record and marks it In progress in Airtable to reduce duplicate processing.
3. Uses Tavily to research the main topic, then uses the configured OpenAI model generate a newsletter title and exactly five distinct section topics.
4. For each section, Tavily gathers sources, an OpenAI agent writes one HTML section with inline citations, and another OpenAI agent produces a structured illustration prompt.
5. Uses Replicate as the included image-generation service, polls until completion if needed, then stores the image in Cloudinary and pairs it with its section. Both integrations can be replaced by adapting the relevant nodes and output mappings.
6. Assembles the five illustrated sections into a plain HTML email, generates a subject and preview text, and creates an unscheduled draft broadcast in Kit targeting the Airtable-specified tag audience.
7. Records the Kit Broadcast ID back to Airtable as Draft and, if enabled, schedules the same broadcast in Kit and updates Airtable with Scheduled status and the scheduled send time.

Setup

1. Create an Airtable base/table with fields for Newsletter Topic, Audience Tag ID (numeric Kit tag ID), Status, Broadcast ID, and Scheduled Date, then add Airtable credentials and set the base/table IDs in the configuration.
2. Configure credentials for the included services: Tavily (Authorization: Bearer …), Replicate (Authorization: Bearer …), and Kit (X-Kit-Api-Key: …). If replacing a service, adapt its request and authentication settings.
3. Add OpenAI credentials and select available text and image-prompt models. To use another provider, replace the relevant AI nodes and update their output mappings where necessary.
4. Configure the included Cloudinary upload using a restricted unsigned preset, cloud name, and preset name. Alternatively, replace it with ImageKit, Google Drive, S3-compatible storage, or your preferred service. Map a direct image URL that email clients can access without signing in, and verify rendering in a test email.
5. Set sender name, contact email, and postal address in the configuration, and keep deliveryMode=draft until you verify the Kit draft content and audience filtering.

Requirements

• An n8n instance with the node versions used in the template.
• An Airtable base with Newsletter Topic, Audience Tag ID, Status, Broadcast ID, and Scheduled Date fields. Setup notes list the required field types and status options.
• OpenAI API access for the included AI nodes.
• Tavily API access for topic and section research.
• Access to an image-generation service. Replicate is configured as the example, but users can replace it with their preferred provider by adapting the image-generation request, authentication, polling if needed, and output mapping.
• Image storage or hosting: Cloudinary is preconfigured. You can replace it with ImageKit, Google Drive, S3-compatible storage, or your preferred service by adapting the upload step and returned image URL.
• Kit API access and an existing audience tag with a numeric ID.
• Sender name, contact email, and postal address for the email footer.

Customization

• Adjust the schedule, newsletter audience, and writing tone.
• Choose available text and image-direction models in Configure Newsletter.
• Replace the image-storage integration with your preferred provider. Map its returned URL into the section-image step. Newsletter images must be accessible to recipients’ email clients without authentication.
• Configure your Airtable base/table IDs and each newsletter’s Kit audience tag ID.
• Keep deliveryMode: draft for review in Kit, or select schedule and configure the sending delay after testing.
• Adapt the plain HTML layout and section prompts to your needs.

Additional info

• Produces five researched sections, each with a generated illustration and source citations.
• Defaults to an unscheduled, non-public Kit draft.
• Optional scheduling records the scheduled time; it does not confirm delivery to subscribers.
• Includes configuration instructions, recovery guidance, and a testing checklist on the canvas.
• AI-generated claims, image text, and illustrations require review.
• Avoid overlapping executions. Failed records require manual inspection before retrying.
• Research, model, image-generation, and hosting costs depend on usage.
• Uses built-in nodes; no private renderer or local file processing is required.