Introducing Ketchup Skills

The biggest friction in using Ketchup has been the context switch. You write code in VS Code or Cursor, but you have to go to a web dashboard to see your team's velocity or generating a changelog.

Today, we are deleting that friction.

Introducing Ketchup Skills: A Model Context Protocol (MCP) server that brings the power of the Narrative Engine directly into your AI workflow.

What can you do?

Once installed, your AI assistant (Claude, Cursor, etc.) gains new specialized tools:

1. Code Forensics: analyze_complexity(path)

   • Ask Claude: "Why is this module so hard to test?"
   • It runs a cyclomatic complexity audit locally and visualizes the hotspots.

2. Momentum Scoring: get_momentum(commits)

   • Ask Claude: "Are we moving faster or slower than last week?"
   • It calculates your team's velocity score, factoring in "Durable Code" vs churn.

3. Automated Recaps: generate_recap(since)

   • Ask Claude: "Write a release note for the last 24 hours of work."
   • It parses your git history, filters the noise, and writes a perfect summary.

Local First, Privacy Always

The best part? It runs locally.

Most developer tools require you to upload your repo to their cloud. Ketchup Skills run 100% on your machine.

• complexity analysis? Local.
• Security scanning? Local.
• AI pattern detection? Local.

Your source code never leaves your device unless you explicitly ask to generate a hosted video.

The Hybrid Video Engine

Speaking of video, we are also unbundling our video engine.

We've introduced a Tiered Architecture for video generation:

• Free Tier (Remotion): Generate clean, professional motion graphics updates using React code. It's fast, deterministic, and free.
• Pro Tier (FAL.ai): Upgrade to get generative AI video (via Kling or MiniMax) for that "Superbowl Commercial" look.

The new MCP server can orchestrate both. You can ask Claude: "Generate a 30s video script for this release," and then pipe that script directly into our rendering engine.

How to Install

If you are using Claude Desktop:

npx skills add gitketchup/ketchup-skills --skill ketchup-changelog

Or allow it to run via uvx or pip:

{
  "mcpServers": {
    "ketchup": {
      "command": "uvx",
      "args": ["ketchup-mcp"]
    }
  }
}

Why this matters

We believe the future of tooling isn't "another dashboard." It's "tools that live where you work." By adopting the Model Context Protocol, Ketchup becomes a native extension of your intelligence, not just another tab in your browser.

Try it out and let us know what you build.