https://github.com/thaolst/ai-growth-agents-for-marketers
AI agents for growth marketing — built from real fintech campaigns. Prompts + Python.
https://github.com/thaolst/ai-growth-agents-for-marketers
ai-agents campaign-optimization claude fintech growth-marketing llm marketing-automation mobile-payment prompt-engineering sea-fintech southeast-asia vietnam
Last synced: 24 days ago
JSON representation
AI agents for growth marketing — built from real fintech campaigns. Prompts + Python.
- Host: GitHub
- URL: https://github.com/thaolst/ai-growth-agents-for-marketers
- Owner: thaolst
- License: mit
- Created: 2026-05-23T03:12:21.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2026-05-30T16:11:09.000Z (about 1 month ago)
- Last Synced: 2026-05-30T17:17:36.407Z (about 1 month ago)
- Topics: ai-agents, campaign-optimization, claude, fintech, growth-marketing, llm, marketing-automation, mobile-payment, prompt-engineering, sea-fintech, southeast-asia, vietnam
- Language: Python
- Homepage: https://www.linkedin.com/in/thaolst/
- Size: 98.6 KB
- Stars: 3
- Watchers: 0
- Forks: 1
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
Awesome Lists containing this project
README
🌐 Website ·
💬 Discussions ·
📝 Contribute ·
📋 Changelog
## Cài đặt qua Agent Skills
```bash
npx skills add thaolst/ai-growth-agents-for-marketers
```
Hoạt động với Claude Code, OpenAI Codex, Cursor, và mọi agent hỗ trợ Agent Skills spec.
> Không phải template. Không phải lý thuyết. Agent thật, chạy được, từ campaign thực tế.
> Not templates. Not theory. Real agents, actually run, from live campaigns.
# Tiếng Việt
## AI Growth Agents for Marketers
Mình dùng AI agent trong công việc growth marketing hàng ngày - để viết brief, phân tích A/B test, lên campaign plan, đề xuất mechanic voucher. Repo này là những gì thực sự chạy được, không phải những gì trông hay trên slide.
### Kết quả thực tế
| Công việc | Trước | Sau |
|----------|-------|-----|
| Viết campaign brief | 3 tiếng | 20 phút |
| Phân tích A/B test | 3 tiếng | 15 phút |
| Báo cáo growth hàng tuần | 4 tiếng | Tự động |
| Đề xuất voucher mechanic | Cảm tính | 10 phút |
### Ai nên dùng
- Growth marketer muốn tự động hóa công việc lặp lại
- Marketing manager cần brief / plan / report nhanh
- Founder / solo marketer làm nhiều việc một lúc
- Bất kỳ ai làm fintech, e-commerce, super app ở Đông Nam Á
### Có gì trong này
| Agent | Làm gì | Cần code? |
|-------|--------|-----------|
| [💡 AI Agent Consultant](./01-what-is-ai-agent/) | Tư vấn chọn agent phù hợp | ❌ |
| [📋 Campaign Brief Agent](./02-your-first-campaign-agent/) | 5 input → brief hoàn chỉnh | ❌ |
| [🔍 Campaign Synthesis](./03-tool-use-no-code/) | Tổng hợp nhiều file campaign | ❌ |
| [⚙️ Automation Scripter](./04-tool-use-python/) | Tự viết Python script | 🐍 |
| [🔄 Multi-Agent Pipeline](./05-multi-agent-workflow/) | Research → Planning | 🐍 |
| [📚 Agentic RAG](./06-agentic-rag/) | Hỏi đáp từ campaign history | ❌ |
| [🎯 Planning Agent](./07-planning-agent/) | Target → Campaign plan | 🐍 |
| [📊 A/B Test Analyzer](./08-ab-test-analyzer/) | Phân tích + đề xuất | 🐍 |
| [📈 MEU Planning](./09-meu-planning-agent/) | MEU target → Campaign plan | 🐍 |
| [🚀 Production Review](./10-agents-in-production/) | Review agent trước deploy | 🐍 |
| [📖 Case Studies](./case-studies/) | Kết quả thực tế, ẩn danh | - |
### Bắt đầu
**Không code - 1 phút:**
```bash
# Mở Claude.ai → paste prompt từ bất kỳ folder 01, 02, 03, 06
```
**Có Python - 2 phút:**
```bash
git clone https://github.com/thaolst/ai-growth-agents-for-marketers
cd ai-growth-agents-for-marketers
make setup
# Edit .env → dán API key → make run AGENT=08-ab-test-analyzer
```
**One-click - 0 phút setup:**
[](https://codespaces.new/thaolst/ai-growth-agents-for-marketers)
Môi trường Python + extensions sẵn, chỉ cần thêm API key.
### Cấu trúc mỗi agent
```
📁 agent-name/
├── README.md ← Dùng nó như thế nào
├── prompt.md ← Prompt thật, copy-paste và dùng
├── agent.py ← Python (nếu cần tự động chạy)
├── requirements.txt ← Dependencies (nếu có Python)
└── example-output.md ← Output thật từ campaign thật
```
### 🌟 Điều làm repo này khác biệt
| Không giống ai | Vì sao độc đáo |
|----------------|----------------|
| **MEU Planning agent** | Repo duy nhất có agent lập kế hoạch ngược từ MEU - phổ biến ở fintech SEA, không ai làm |
| **Python agents chạy thật** | Skills markdown + 7 Python agents executable - dual format |
| **Fintech + SEA context** | Không lý thuyết SaaS Mỹ. Campaign thật ở Việt Nam, Indonesia, Philippines |
| **Data connector** | growth-mcp MCP server → real metrics, không guesswork |
| **Context system** | Điền `.agents/product-marketing-context.md` một lần → 15 skills dùng chung |
| **Bilingual VI/EN** | Prompt và skills 100% song ngữ |
### Agent Skills (dùng với Claude Code, Cursor, Codex)
```bash
# Cài tất cả (15 skills)
npx skills add thaolst/ai-growth-agents-for-marketers
# Cài skill cụ thể
npx skills add thaolst/ai-growth-agents-for-marketers/skills/campaign-brief
```
| Skill | Agent | Mô tả | Context? |
|-------|-------|-------|----------|
| `ai-agent-consultant` | [01](./01-what-is-ai-agent/) | Tư vấn chọn agent phù hợp | ✅ |
| `campaign-brief` | [02](./02-your-first-campaign-agent/) | Viết campaign brief | ✅ |
| `campaign-synthesis` | [03](./03-tool-use-no-code/) | Tổng hợp campaign | ✅ |
| `automation-scripter` | [04](./04-tool-use-python/) | Tạo Python script | ✅ |
| `multi-agent-research` | [05](./05-multi-agent-workflow/) | Research→Plan pipeline | ✅ |
| `rag-knowledge-base` | [06](./06-agentic-rag/) | RAG knowledge base | ✅ |
| `campaign-planning` | [07](./07-planning-agent/) | Campaign planning | ✅ |
| `ab-test-analyzer` | [08](./08-ab-test-analyzer/) | A/B test analysis | ✅ |
| `meu-planning` | [09](./09-meu-planning-agent/) | MEU planning | ✅ |
| `agent-pre-deploy-review` | [10](./10-agents-in-production/) | Pre-deploy review | ✅ |
| `voucher-mechanic-designer` | - | Thiết kế mechanic voucher | ✅ |
| `fintech-campaign-designer` | - | Campaign design cho fintech | ✅ |
| `retention-analyzer` | - | Phân tích retention | ✅ |
| `churn-intervention` | - | Save offer + winback | ✅ |
| `growth-mcp-connect` | - | **Data connector** → real data | 🟢 |
> ✅ = Skill tự động đọc `.agents/product-marketing-context.md`
> 🟢 = Skill kết nối growth-mcp để lấy real-time data
### 💡 Hệ thống Context (mới)
```bash
.agents/
├── product-marketing-context.md ← Điền 1 lần → 15 skills tự động dùng
└── growth-metrics-context.md ← Tùy chọn: metric baseline thủ công
```
Cách này giống Corey Haines (31k⭐ marketing skills) nhưng fintech-focused + có Python agents.
### 🔌 Data Connector: growth-mcp (mới)
Kết nối growth-mcp để pull real metrics:
```bash
git clone https://github.com/thaolst/growth-mcp.git
cd growth-mcp && pip install -e .
```
Khi growth-mcp chạy, các skill sẽ tự động:
- Pull campaign performance trước khi viết brief
- Pull real cohort data trước khi phân tích retention
- Pull experiment data trước khi phân tích A/B test
- So sánh recommendation với campaign history thật
### Chọn provider
Mặc định dùng Anthropic Claude. Set `AI_PROVIDER=openai` trong `.env` để dùng OpenAI GPT.
```bash
# Mặc định (Anthropic)
AI_PROVIDER=anthropic
ANTHROPIC_API_KEY=sk-ant-...
# Hoặc OpenAI
AI_PROVIDER=openai
OPENAI_API_KEY=sk-proj-...
```
| Biến | Mặc định | Ghi chú |
|------|----------|---------|
| `AI_PROVIDER` | `anthropic` | `anthropic` hoặc `openai` |
| `ANTHROPIC_API_KEY` | - | [console.anthropic.com](https://console.anthropic.com) |
| `OPENAI_API_KEY` | - | Chỉ cần nếu dùng OpenAI |
| `AI_MODEL` | (default) | Ghi đè model nếu muốn |
### Hệ sinh thái repo
Mình có 3 repo phục vụ 3 mục đích khác nhau:
| Repo | Là gì | Dùng khi nào |
|---|---|---|
| [ai-growth-prompts](https://github.com/thaolst/ai-growth-prompts) | Thư viện prompt theo chủ đề, copy-paste được ngay | Cần prompt cho 1 task cụ thể: thiết kế voucher, phân tích segment, viết brief |
| [ai-growth-agents-for-marketers](https://github.com/thaolst/ai-growth-agents-for-marketers) (repo này) | Workflow nhiều bước dạng prompt + script, có skill cài cho Claude Code | Muốn chạy quy trình end-to-end: lập kế hoạch MEU, phân tích A/B test |
| [growth-mcp](https://github.com/thaolst/growth-mcp) | MCP server đóng gói logic growth thành tool | Muốn Claude/Cursor gọi tool trực tiếp thay vì paste prompt |
### Tham gia cộng đồng
- [💬 Discussions](https://github.com/thaolst/ai-growth-agents-for-marketers/discussions) - hỏi đáp, chia sẻ kinh nghiệm
- [🐛 Report bug](https://github.com/thaolst/ai-growth-agents-for-marketers/issues/new?template=bug_report.md) - prompt không hoạt động
- [💡 Suggest agent](https://github.com/thaolst/ai-growth-agents-for-marketers/issues/new?template=feature_request.md) - đề xuất agent mới
- [📝 Pull Request](./CONTRIBUTING.md) - đóng góp prompt/agent
### Tác giả
**Lê Song Tiên Thảo** - Growth Marketer @ MoMo (fintech, SEA)
Mình build agent, prompt, và workflow cho công việc growth thực tế - trong giới hạn thật của budget, channel, và segment behavior.
[LinkedIn](https://www.linkedin.com/in/thaolst/) · [GitHub](https://github.com/thaolst) · [Substack](https://thaolst.substack.com/) · [Website](https://thaolst.github.io/ai-growth-agents-for-marketers)
⭐ Star repo này nếu nó tiết kiệm cho bạn ít nhất 1 tiếng. ⭐
# English
## AI Growth Agents for Marketers
I use AI agents daily for growth marketing work - writing briefs, analyzing A/B tests, planning campaigns, recommending voucher mechanics. This repo is what actually runs, not what looks good on slides.
### Real results
| Task | Before | After |
|------|--------|-------|
| Campaign brief | 3 hours | 20 minutes |
| A/B test analysis | 3 hours | 15 minutes |
| Weekly growth report | 4 hours | Automated |
| Voucher recommendation | Gut feeling | 10 minutes |
### Who uses this
- Growth marketers who write the same brief structure every week and want that part automated
- Marketing managers running on tight timelines - brief, plan, report, repeat
- Founders doing growth alone without a team
- Anyone running campaigns in fintech, e-commerce, or super apps across SEA
### What's inside
| Agent | What it does | Code? | Context? |
|-------|-------------|-------|----------|
| [💡 AI Agent Consultant](./01-what-is-ai-agent/) | Choose the right agent type | ❌ | ✅ |
| [📋 Campaign Brief Agent](./02-your-first-campaign-agent/) | 5 inputs → complete brief | ❌ | ✅ |
| [🔍 Campaign Synthesis](./03-tool-use-no-code/) | Synthesize multiple files | ❌ | ✅ |
| [⚙️ Automation Scripter](./04-tool-use-python/) | Auto-generate Python scripts | 🐍 | ✅ |
| [🔄 Multi-Agent Pipeline](./05-multi-agent-workflow/) | Research → Planning | 🐍 | ✅ |
| [📚 Agentic RAG](./06-agentic-rag/) | Q&A from campaign history | ❌ | ✅ |
| [🎯 Planning Agent](./07-planning-agent/) | Target → full campaign plan | 🐍 | ✅ |
| [📊 A/B Test Analyzer](./08-ab-test-analyzer/) | Analyze + recommend | 🐍 | ✅ |
| [📈 MEU Planning](./09-meu-planning-agent/) | MEU target → campaigns | 🐍 | ✅ |
| [🚀 Production Review](./10-agents-in-production/) | Pre-deploy review | 🐍 | ✅ |
| [💰 Voucher Mechanic Designer](#) | Voucher/cashback design | ❌ | ✅ |
| [🏦 Fintech Campaign Designer](#) | SEA fintech campaign design | ❌ | ✅ |
| [📉 Retention Analyzer](#) | Cohort retention diagnosis | ❌ | ✅ |
| [🛑 Churn Intervention](#) | Save offers + winback | ❌ | ✅ |
| [🔌 growth-mcp Connect](#) | **Data connector** → live data | 🐍 | 🟢 |
| [📖 Case Studies](./case-studies/) | Anonymized real results | - | - |
> ✅ = Auto-reads `.agents/product-marketing-context.md` for shared context
> 🟢 = Connects to growth-mcp MCP server for real-time data
### 🎯 What Makes This Different
Unlike every other marketing agent repo (Corey Haines, LeoYeAI, AgentKits), this one:
| Unique | Why it matters |
|--------|---------------|
| **MEU Planning agent** | Only repo with MEU-backward campaign planning - common in SEA fintech, ignored everywhere else |
| **Dual format** | Skills + executable Python agents - not just markdown |
| **Fintech + SEA** | Real campaign context from Vietnam, Indonesia, Philippines |
| **Data connector** | growth-mcp MCP server → live metrics, not guesswork |
| **Context system** | Fill `.agents/product-marketing-context.md` once → 15 skills share it |
| **Bilingual VI/EN** | Full Vietnamese + English prompts and skills |
### 💡 Context System
```
.agents/
├── product-marketing-context.md ← Fill once → 15 skills auto-use it
└── growth-metrics-context.md ← Optional: manual metric overrides
```
Inspired by Corey Haines (31k⭐) but fintech-focused + with real Python agents.
### 🔌 Data Connector: growth-mcp
Connect [growth-mcp](https://github.com/thaolst/growth-mcp) for live campaign data:
```bash
git clone https://github.com/thaolst/growth-mcp.git
cd growth-mcp && pip install -e .
```
When connected, skills automatically:
- Pull campaign performance before writing briefs
- Pull real cohort data before retention analysis
- Pull experiment data before A/B test analysis
- Compare recommendations against real campaign history
### Quick start
**No code - 1 minute:**
```bash
# Open Claude.ai → paste any prompt from folders 01, 02, 03, 06
```
**Python - 2 minutes:**
```bash
git clone https://github.com/thaolst/ai-growth-agents-for-marketers
cd ai-growth-agents-for-marketers
make setup
# Edit .env → paste API key → make run AGENT=08-ab-test-analyzer
```
**One-click - 0 setup:**
[](https://codespaces.new/thaolst/ai-growth-agents-for-marketers)
### Provider support
All Python agents support both **Anthropic Claude** (default) and **OpenAI GPT**. Set `AI_PROVIDER=openai` in `.env` to switch, or override with `AI_MODEL`.
### Repo ecosystem
I maintain 3 repos serving different purposes:
| Repo | What it is | When to use |
|---|---|---|
| [ai-growth-prompts](https://github.com/thaolst/ai-growth-prompts) | Topic-based prompt library, ready to copy-paste | You need a prompt for one specific task: voucher design, segment analysis, campaign brief |
| [ai-growth-agents-for-marketers](https://github.com/thaolst/ai-growth-agents-for-marketers) (this repo) | Multi-step workflows as prompts + scripts, installable as Claude Code skills | You want an end-to-end process: MEU planning, A/B test analysis |
| [growth-mcp](https://github.com/thaolst/growth-mcp) | MCP server packaging growth logic as callable tools | You want Claude/Cursor to call tools directly instead of pasting prompts |
### Join the community
- [💬 Discussions](https://github.com/thaolst/ai-growth-agents-for-marketers/discussions) - Q&A, share experiences
- [🐛 Report bug](https://github.com/thaolst/ai-growth-agents-for-marketers/issues/new?template=bug_report.md) - prompt not working
- [💡 Suggest agent](https://github.com/thaolst/ai-growth-agents-for-marketers/issues/new?template=feature_request.md) - suggest a new agent
- [📝 Pull Request](./CONTRIBUTING.md) - contribute prompts/agents
### Author
**Lê Song Tiên Thảo (Tara)** - Growth Marketing Manager @ MoMo (fintech, SEA)
Building AI agents, prompts, and workflows for real growth work - within real constraints of budget, channel, and segment behavior.
[LinkedIn](https://www.linkedin.com/in/thaolst/) · [GitHub](https://github.com/thaolst) · [Substack](https://thaolst.substack.com/) · [Website](https://thaolst.github.io/ai-growth-agents-for-marketers)
⭐ Star this if it saves you at least 1 hour. ⭐