Blog and content API

Cauliflower includes a small headless CMS for your company blog. Platform admins write posts in the admin portal; they're published at /blog with an RSS feed, a sitemap and social previews — and every post is also available over a JSON API, so agents, scripts or another website can read and write content.

Writing posts

Open Admin → Content → New post:

• Write in Markdown (headings, lists, tables, code, quotes and images) with a side-by-side preview that renders exactly like the blog. Upload images straight from the toolbar.
• Publishing — leave the publish date empty to publish now, or pick a future date to schedule the post. Set the URL, tags and author.
• Cover image — 1200×630 works best; it's used on the blog and as the link preview. Posts without a PNG or JPEG cover get a generated preview with their title.
• Search & sharing — meta title and description with a search result preview, and a canonical URL when the post first appeared elsewhere.
• Social posts — LinkedIn and X copy for promoting the post. Write {url} where the post's link goes. Draft with AI writes both from the article using the provider configured under Admin → AI assistant.

Unpublish turns a post back into a draft. Every change is recorded in the platform audit log.

Sharing posts automatically

Under Admin → Content → Publishing & API, set a publish webhook URL. When a post goes live — immediately, or when its scheduled time arrives — Cauliflower sends one POST:

{
  "event": "post.published",
  "post": {
    "id": "post_…", "slug": "why-demo-requests-go-cold", "url": "https://cal.example.com/blog/why-demo-requests-go-cold",
    "title": "…", "excerpt": "…", "cover_image_url": "…", "tags": ["speed-to-lead"],
    "published_at": "2026-10-08T09:00:00.000Z",
    "social": { "linkedin": "…", "x": "…" }
  }
}

Point it at Zapier, Make, n8n or your own endpoint to queue the social posts in Buffer, post to LinkedIn or Slack, or send a newsletter. The RSS feed at /blog/rss.xml works with newsletter tools and feed readers too.

Content API

Base URL: https://cal.example.com/api/content

Reading (public)

Published posts can be read without authentication, from any origin (CORS is open for GET):

Each post has id, slug, url, title, excerpt, content_markdown, cover_image_url, author, tags, status, published_at, updated_at, reading_minutes, seo and social.

Writing (token)

Create a token under Admin → Content → Publishing & API. It's shown once; send it as Authorization: Bearer cms_….

curl -X POST https://cal.example.com/api/content/posts \
  -H "Authorization: Bearer $CMS_TOKEN" -H "Content-Type: application/json" \
  -d '{"title":"Routing leads by territory","content":"## Why territories…","tags":["routing"]}'

Every write response includes images: how many external images were copied and a warning for any that couldn't be.

Images

External images in a post's Markdown (![alt](https://…) and <img> tags) and its cover_image_url are copied into Cauliflower when you create or update the post, so posts never hotlink other sites or break when those images move. Images that can't be downloaded keep their link and are listed in images.warnings. Send "rehost_images": false to skip this.

To upload an image yourself, POST /media with a CMS token and one of:

It returns { "url": "…", "data": [{ "id", "url", "content_type", "size" }] }. Use the url as cover_image_url or in Markdown. Images can be PNG, JPEG, WebP, GIF or AVIF up to 5 MB; SVG isn't accepted. Downloads only reach public addresses, never private networks.

Brief and schema

Tokens can be revoked at any time. A good pattern is to let an AI agent create drafts on a schedule and have a person review and publish them.

Growth agent

Admin → Content → Growth agent turns ChatGPT (or any agent that can make HTTP requests) into a writer that runs the blog on its own. Each run, it reads the brief, picks a topic that no post covers yet, researches it with web search, writes the article in your voice with internal and external links, makes a cover and in-body illustrations (or finds licensed photos), uploads them, writes the SEO fields and social copy, and posts it as a draft (or publishes it, if you choose).

1. Fill in What the agent knows: the product, audience, voice, topics, hard rules, image style and call to action. Empty fields use sensible defaults. The agent reads these from the brief every run, so you can change them later without touching the prompt.
2. Click Put a new token in the prompt and Copy prompt. The prompt stays under ChatGPT's 8,000-character limit for GPT instructions.
3. In ChatGPT, create a GPT: paste the prompt into Instructions, turn on Web Search and Image Generation, add an action with Import from URL set to https://cal.example.com/api/content/openapi.json, and set its authentication to API Key → Bearer with the token.
4. Say "Write the next post" whenever you want one, or ask ChatGPT to schedule it.

Drafts appear under Content for review. Publishing still triggers the publish webhook, so social posts go out the same way as posts you write yourself.