From 3dplotter Code to 415 Scheduled Posts: MCP in Cursor

How we pointed Reaudit MCP at the real 3dplotter feature manifests, generated platform-native copy from codebase facts, and queued 152 days of posts without opening a scheduler UI.

Where Parts 1–6 left off

The first six articles in this series documented how 3dplotter.xyz reached a 93/100 AI Visibility Score, 65,000 citations across ChatGPT, Perplexity, and Google AI, and $11,926 in AI-attributed revenue with zero paid ads. The growth engine was Reaudit's GTM strategy, MCP via your harness (Cursor, Grok bot, Claude, Hermes, or any other client), the content factory, prompt tracking, and Stripe revenue attribution.

Part 6 ended with Broadcast wired up but not yet activated. By August 2026 the product had citations and sales, but social distribution was barely moving. 3dplotter already had accounts on YouTube, Facebook, Instagram, TikTok, and X, yet the August 23 baseline showed 10 posts, zero views, and two likes. Signups came from AI referrals and search engines; off-site content was mostly six articles on Instructables. Social was the missing layer: not another scheduler tab, but an agent that could read the product codebase and queue months of posts through Broadcast.

The baseline on August 23

Before the V3 campaign opened, we captured an honest baseline from the 3dplotter social accounts (stored in v3-social-launch-manifest.json as baselineAnalytics):

That is the starting point for everything that follows. The claim in this article is not that social caused the AI visibility score. It is that an agent orchestrated 152 days of grounded content from the codebase and queued it through Broadcast.

Two campaigns, one queue

The V3 social launch is two campaigns scheduled back-to-back through the same Broadcast queue:

Campaign A: 14-day launch run (Aug 24 – Sep 6)

3DPlotter V3: 14-Day Social Launch covers the V3 product launch narrative: tease, problem, studio reveal, generators, modifiers, Genplot AI, full workflow, countdown, and launch day. Each day has a theme, a long-form article, and social posts scheduled across all six platforms. 70 posts were queued through Reaudit MCP; 65 were independently verified on the content calendar.

Campaign B: Feature series (Sep 8 – Jan 22)

3DPlotter V3: Feature Series follows every-other-day cadence (2-day gap) at 07:00 UTC (10:00 Athens). One post per product feature: 69 features (48 objects + 21 modifiers), each fanned out to six platforms = 345 social slots.

Combined: 415 social slots across 152 days (Aug 24, 2026 – Jan 22, 2027), plus 14 long-form launch articles generated to back the first two weeks.

The agent reads 3dplotter, not a brief

We opened the 3dplotter repo with Reaudit MCP connected in our harness (Cursor in this run). Instead of a marketing brief, the agent reads manifests under lib/marketing/: feature metadata, canonical URLs, menu paths, SEO keywords, and platform copy slots. Every post cites something checkable, like a menu path Add Object → Shapes & Imports → Shape, or a count like 66 built-in shapes.

Manifests the agent actually used

These files live in the 3dplotter.xyz repository. The MCP workflow reads them on every feature post so copy stays grounded in product facts rather than generic AI filler.

export interface V3FeaturePost {
  readonly postNumber: number;
  readonly label: string;
  readonly group: string;
  readonly canonicalUrl: string;
  readonly menuPath: string;
  readonly codebaseFact: string;
}

/** 69 features: 48 objects + 21 modifiers — sourced from the product menu tree */
export const V3_FEATURE_POSTS: readonly V3FeaturePost[] = [
  {
    postNumber: 1,
    label: 'Shape',
    group: 'Shapes & Imports',
    canonicalUrl: 'https://3dplotter.xyz/features/shape',
    menuPath: 'Add Object → Shapes & Imports → Shape',
    codebaseFact:
      '66 built-in shapes across 8 categories: Lines, Rectangles, Basic Shapes, Arrows, Equation, Flowchart, Stars & Banners, and Callouts.',
  },
  // … 68 more features, each grounded in lib/plotter/objects/*
];

{
  "campaign": "3dplotter-v3-feature-series",
  "platforms": ["instagram", "tiktok", "linkedin", "twitter", "facebook", "youtube"],
  "cadenceDays": 2,
  "postTimeUtc": "07:00",
  "posts": [
    {
      "postNumber": 1,
      "feature": "shape",
      "group": "Shapes & Imports",
      "scheduledAt": "2026-09-08T07:00:00Z",
      "canonicalUrl": "https://3dplotter.xyz/features/shape",
      "menuPath": "Add Object → Shapes & Imports → Shape",
      "copy": {
        "twitter": "Shape — 66 built-in shapes: lines, rectangles, polygons, arrows, flowchart, stars, and callouts. Add Object → Shapes & Imports → Shape. https://3dplotter.xyz/features/shape #3DPlotter #PenPlotter",
        "linkedin": "V3 Feature 1/69: Shape\n\n**From the 3DPlotter codebase:** 66 built-in shapes…"
      },
      "assets": {
        "image": "/marketing/v3/shape-hero.png",
        "video": "/marketing/v3/shape-demo.mp4"
      },
      "seo": {
        "title": "Shape tool — 66 built-in plotter shapes | 3DPlotter V3",
        "keywords": ["pen plotter shapes", "3dplotter shape tool", …]
      }
    }
  ]
}

{
  "campaign": "3dplotter-v3-14-day-launch",
  "startDate": "2026-08-24",
  "days": 14,
  "scheduledPosts": 70,
  "baselineAnalytics": {
    "capturedAt": "2026-08-23",
    "posts": 10,
    "views": 0,
    "likes": 2
  },
  "themes": [
    { "day": 1, "date": "2026-08-24", "theme": "The Tease" },
    { "day": 14, "date": "2026-09-06", "theme": "LAUNCH DAY" }
  ]
}

What each feature post carries

Per the manifest schema, every one of the 69 feature posts includes:

• Platform-native copy for all 6 channels
• SEO meta title + description
• 12 target keywords
• Canonical feature page URL
• Rendered image asset
• Demo video asset

Sample output: Feature 1/69 (Shape)

This is real copy generated for the first feature post, scheduled 2026-09-08, drawn from the Shapes & Imports group. The menu path and object count come straight from the codebase.

Tool chain: write, illustrate, schedule, verify

The agent followed the same sequence documented in the Broadcast campaign scripts. No Buffer. No copying between tabs. Each MCP tool call produces an artifact the next step consumes:

generate_content drafts the long-form feature article with SEO meta and keywords. generate_social_image renders an on-brand image per post. cross_post_to_social fans one draft to Instagram, TikTok, LinkedIn, X, Facebook, and YouTube with platform-native copy. schedule_social_post (or bulk_schedule_social_posts for batches) places each slot at 07:00 UTC. get_content_calendar verifies the queue landed as planned.

Six channels from one queue

Each platform gets copy shaped for its constraints (character limits, hashtag norms, link handling), but every version points back to the same canonical feature URL and codebase facts.

InstagramTikTokLinkedInXFacebookYouTube

Human approval stays in the loop

The agent prepares and queues. Campaign posts still go through human approval before anything publishes, same as the Reaudit dashboard flow. After bulk scheduling, we call get_content_calendar to confirm dates, channels, and copy landed correctly.

What to take away

The point is not that AI wrote 415 posts. It is that every post is traceable to a real feature in the 3dplotter codebase, scheduled through the same system that tracks AI visibility and attributes revenue. You can audit any slot back to a manifest row, a menu path, and a URL on 3dplotter.xyz.

Part 8 covers what happened after Broadcast went live on August 24: measured traffic, signups, and Stripe revenue in equal 19-day windows, with interactive charts and full methodology.