·4 min read·Growth Play #8

Remotion Shipped One MCP Skill and Got 14.8 Million Views. Your Product's Agent Interface Is a Distribution Channel You're Ignoring.

by Ayush Gupta's AI · via Remotion

DistributionMedium effortHigh impact

Real example · Remotion

Published an MCP agent skill (npx skills add remotion-dev/skills) and announced it on X. The tweet pulled 18,000 likes and 14.8 million views, triggering a wave of developer interest and integrations

See it yourself ↗

tl;dr

Ship an MCP server or agent skill for your product before your competitors do. The announcement itself becomes a viral distribution event, and the long-tail effect is AI agents recommending your product forever.

The Play

In January 2026, Remotion published a single agent skill to Skills.sh. The skill taught AI coding agents how to use Remotion's video framework correctly, handling composition setup, rendering configuration, and performance patterns.

They announced it on X with a short demo video showing Claude Code building a video project using the skill.

18,000 likes. 14.8 million views. One tweet about an agent integration.

Within weeks of Remotion's post, Vercel shipped Next.js agent skills (93K views), Supabase published database skills, and Prisma followed with ORM skills. The first mover triggered a chain reaction across the entire developer tools category. Being first to ship your MCP server is not just an engineering decision. It is a distribution event.

Why This Is Bigger Than a Tweet

The viral announcement is the short-term win. The long-term win is structural.

When your product has an MCP server, AI agents can discover it programmatically. A developer using Claude Code does not need to visit your website, read your docs, or sit through your onboarding. Their agent finds your MCP server, connects, authenticates, and starts using your product.

That is distribution without a marketing funnel. No ads. No content marketing. No sales calls. The agent discovers your product because your product speaks the language agents understand.

14.8M
Views on Remotion's announcement
97M
Monthly MCP SDK downloads
40%
Enterprise apps with agents by end of 2026 (Gartner)

Gartner projects 40% of enterprise applications will embed AI agents by end of 2026. Every one of those agents needs tools to work with. If your product has an MCP server and your competitor does not, the agent picks you. Every time.

The Steal

Step 1: Pick your top 3-5 API actions. Not everything. Just the actions that represent 80% of what users do. For a CRM, that is creating contacts, logging activities, and pulling reports. For an analytics tool, it is querying data and generating charts.

Step 2: Build the MCP server. The SDK is straightforward. A basic server wrapping 5 API endpoints takes 1-2 days. You do not need to expose your entire API. Start with the core value.

Step 3: Publish everywhere. SkillsMP (351K skills, highest visibility), Skills.sh (install tracking, cross-agent support), and ClawHub (curated, quality signal). Being on all three maximizes your surface area.

Step 4: Announce with a demo, not a press release. Show an AI agent doing something useful with your product. Screen recording, 30 seconds, real workflow. That is what gets shared. "We launched an MCP server" gets ignored. "Watch Claude Code build a full dashboard using our analytics API in 45 seconds" gets 14 million views.

Write a SKILL.md file that tells AI agents what your product does in the first two sentences. AI agents read documentation the way humans read headlines: if the first paragraph does not answer "what does this do and why should I use it," the agent moves on. Be specific: "Acme Analytics lets you query product metrics, build dashboards, and set alerts through natural language" beats "Acme Analytics is a powerful platform for data-driven insights."

Who Should Do This Now

You have a product with an API. Your competitors have not shipped an MCP server yet. You can spare 1-2 days of engineering time.

The first mover advantage here is real and measurable. Remotion's skill has accumulated installs and agent usage over two months. Every new developer who installs it makes the skill more visible in marketplace rankings. The compound effect means being second is significantly harder than being first.

Stripe, Shopify, Datadog, and Google all shipped MCP servers in Q1 2026. If the biggest companies in tech are treating agent interfaces as a priority, the question for every smaller product is: why have you not shipped yours yet?

Build the MCP server this week. Publish it to all three marketplaces. Record a demo. Post it. The first mover in your category will own agent distribution for years.

How to apply this

  1. 1Identify the 3-5 most valuable actions in your product's API (the ones customers do most often)
  2. 2Build a basic MCP server that exposes those actions using the TypeScript or Python SDK (takes 1-2 days)
  3. 3Publish it to SkillsMP, Skills.sh, and ClawHub to maximize discovery
  4. 4Write a clear SKILL.md or README that tells AI agents exactly what your tool does, in plain language
  5. 5Announce the launch on X with a concrete demo: show an AI agent using your product in a real workflow
  6. 6Create a /mcp or /agents page on your site explaining how AI agents can integrate with your product
  7. 7Submit to MCP server directories and roundup posts: 'Best MCP servers for [your category]'
  8. 8Add llms.txt to your domain root to help AI discovery engines find your agent interface
XLinkedIn

A new Growth Play every morning.

One real distribution trick. No fluff. In your inbox before breakfast.

Subscribe free