# Claude Code for Growth Marketing

> Anthropic's internal framework for agentic marketing automation using Claude Code

A breakdown of how a solo growth marketer at Anthropic uses Claude Code to handle tasks typically requiring engineering support. It details practical agentic workflows for Google Ads generation, automated Figma design variations, and Meta Ads MCP analytics.

## The framework

The Growth Marketing team (a non-technical team of one) builds performance marketing channels across paid search, paid social, mobile app stores, email marketing, and SEO, using Claude Code to automate repetitive tasks and create agentic workflows.

Main Claude Code use cases:

1. Automated Google Ads creative generation — agentic workflow processes CSVs of hundreds of ads with performance metrics, identifies underperforming ads, generates new variations meeting character limits (30 headlines / 90 descriptions), using two sub-agents (headlines, descriptions). Hundreds of new ads in minutes.

2. Figma plugin for mass creative production — identifies frames and programmatically generates up to 100 ad variations by swapping headlines/descriptions; what took hours of copy-pasting now takes half a second per batch, enabling 10x creative output.

3. Meta Ads MCP server for campaign analytics — MCP server integrated with Meta Ads API to query campaign performance, spending, and ad effectiveness directly in Claude Desktop.

4. Advanced prompt engineering with memory systems — rudimentary memory system logging hypotheses and experiments across ad iterations, pulling previous test results into context for a self-improving testing framework.

Team impact:

- Dramatic time savings on repetitive tasks: ad copy creation reduced from 2 hours to 15 minutes.

- 10x increase in creative output: vastly more ad variations across channels with automated generation and Figma integration.

- Operating like a larger team: handling tasks traditionally requiring dedicated engineering resources.

- Strategic focus shift: more time on strategy and building agentic automation rather than manual execution.

Top tips from the Growth Marketing team:

1. Identify API-enabled repetitive tasks — look for workflows with repetitive actions in tools with APIs (ad platforms, design tools, analytics platforms); prime candidates for automation.

2. Break complex workflows into specialized sub-agents — separate agents per task (headline agent vs description agent) makes debugging easier and improves output quality.

3. Thoroughly brainstorm and prompt plan before coding — spend time upfront using Claude.ai to think through the entire workflow, have Claude.ai create a comprehensive prompt and code structure, and work step-by-step rather than one-shot to avoid Claude getting overwhelmed.

## What it is for

- Generate bulk variations of Google Ads headlines and descriptions from performance CSV data
- Automate mass ad creative production in Figma by programmatically swapping copy
- Query Meta Ads campaign performance directly using a custom MCP server
- Build a persistent memory log of past ad experiments to guide future iterations

## Key points

- Sub-agent workflow design: separating distinct tasks (e.g., headlines vs. descriptions) for better output quality
- Custom Figma plugin generation for instantaneous multi-asset visual rendering
- Model Context Protocol (MCP) server integration to query ad performance data within Claude Desktop
- Iterative testing framework with experiment logging to prevent repeating failed concepts
- Upfront planning methodology: using Claude.ai to outline full architecture before executing in Claude Code

## Pricing

Free

## Details

- Type: Framework
- Category: [Marketing & Sales](https://xiry.app/topics/marketing-sales.md)
- Topic: [Marketing Automation](https://xiry.app/topics/marketing-sales/marketing-automation.md)
- Open source: no
- Platforms: Command line, other
- For: Growth marketers, solo founders, and performance marketers looking to automate repetitive operations
- Tags: claude code, growth marketing, agentic workflows, mcp, google ads, automation

## Why it was saved

- [@robjama, 15 February 2026](https://x.com/robjama/status/2023151595295268939): Concrete examples of agentic marketing automation with measurable impact and practical tips

## Related

- [Claude Superpowers for Marketing Pipeline](https://xiry.app/resources/claude-superpowers-for-marketing-pipeline): A framework using Claude Code and superpowers plugin for marketing workflows (Tip, Free)
- [goose-skills](https://xiry.app/resources/goose-skills): Open-source growth and GTM skills library for AI coding agents (GitHub repo, Freemium)
- [HighLevel](https://xiry.app/resources/highlevel): All-in-one CRM, funnel builder, and marketing automation platform for agencies. (Web app, Paid)
- [SendPulse](https://xiry.app/resources/sendpulse): Multi-channel marketing automation platform with email, CRM, and AI chatbots (Web app, Freemium)
- [Контент Завод Академия](https://xiry.app/resources/kontent-zavod-akademiya): Course on building multi-agent AI pipelines for multi-platform social media distribution (Article or course, Paid)
- [Gooseworks Skills Library](https://xiry.app/resources/gooseworks-skills-library): Catalog of 200+ growth and marketing skills for AI coding agents (Web app, Freemium)

Updated 11 October 2026. Page: https://xiry.app/resources/claude-code-for-growth-marketing

Make your own library: Xiry takes your saved reels, tweets, Reddit bookmarks and Watch Later videos, makes them searchable and pulls the useful stuff out, like this library. It's invite-only for now; join the waitlist at https://xiry.app/#waitlist.