AI Marketing Agents for GTM Engineers

Sources:

• github.com/topics/ai-marketing (113 public repositories, filtered for GTM ICP relevance)
• YouTube: "Marketing Agents for Startups" (Cody Schneider, Agent Native podcast, July 2026)

There are now 113 public repositories on GitHub tagged ai-marketing. Most of them are either generic chatbot wrappers or Claude Code skill files. This playbook filters for the ones that matter to GTM engineers, B2B founders, and outbound-focused marketing teams.

Filtering criteria: repos included must address at least one of these ICP jobs-to-be-done:

• Building or automating outbound (cold email, LinkedIn, paid ads)
• Content production at scale (LinkedIn, social, blog)
• Lead generation and qualification
• Competitive intelligence and market research
• Revenue analytics and attribution

Excluded: generic AI writing assistants, non-English repos, blockchain/crypto marketing, e-commerce only, repos not updated since 2025.

What a real marketing agent is (from Cody Schneider):

Not a Zapier automation workflow that is linear. Not a Claude Code session that runs once. A real marketing agent has three components: (1) a data pipeline and warehouse that gives it unified visibility of the whole pipeline, (2) autonomous decision-making on a cadence with a thinking loop, (3) cloud hosting so it runs continuously and improves based on feedback from live business data.

The full filtered repo list at a glance

Repo	Stars	What it does	ICP relevance
AgriciDaniel/claude-ads	6.2k	250+ checks across Google, Meta, YouTube, LinkedIn, TikTok, Microsoft, Apple Ads with weighted scoring and parallel agents	Paid ads at scale
zubair-trabzada/ai-marketing-claude	1.9k	15 marketing skills with parallel subagents: website audit, copy, email sequences, ad campaigns, content calendars, competitive intel, PDF reports	Full-stack GTM
AgriciDaniel/claude-blog	1.1k	30 sub-skills, 5 agents, 5-gate Blog Delivery Contract (validated brief, draft, quality score, AI detection, fact check), dual-optimised for Google and AI citations	SEO and content
sergebulaev/linkedin-skills	185	Claude Code skills for LinkedIn: human-sounding posts, comments, feed analysis, publishing cadence	LinkedIn growth
indranilbanerjee/digital-marketing-pro	145	158 skills, 25 agents, 12-part strategy flow, 6-platform AEO/GEO including Google AI Mode, EU AI Act ready	Agencies and in-house
stevenflanagan1/social-ai-team	136	Complete AI social team for SMBs: brand setup, content calendar, captions, creative, performance review	Social media
Anil-matcha/Open-AI-UGC	117	Open-source AI UGC video ad generator using Veo 3.1, Seedance 2, Grok Video. Free alternative to Arcads.	Video ads
ScaleBrick/founder-marketing-skills	77	Keyword research, growth strategy, social search audit, competitor analysis. Install with npx skills.	Founders
AgriciDaniel/claude-email	75	Inbox triage, email composition, quality review, deliverability audit, automation sequences, marketing strategy	Email marketing
rediumvex/social-media-caption-generator-claude	73	Captions for Instagram Reels, Carousel, TikTok, Threads, Facebook, YouTube Shorts. Algorithm-optimised for shares, saves, completion rate.	Content creators
AgriciDaniel/linkedin-content-creator	60	Text, image, and carousel post generation with AI research, content calendar, batch generation, direct LinkedIn posting via OAuth	LinkedIn
toofast1/awesome-ai-lead-generation	48	Curated list of AI tools for finding customers, social listening, automated outreach. The modern growth stack.	Lead generation
SaigonXIII/evc	48	Open-source marketing agent workspace for Claude Code: 42 commands, 12 hooks, 4 industry starters	Agent workspace
rediumvex/ai-marketing-claude	46	12 skills, 5 specialist subagents, Python scripts for audits, CRO, SEO, copy, emails, ads, client reports	Agency ops
ericosiu/marketing-os-starter	42	AI marketing team clone: 4 agents, 5 skills, structured handoffs, persistent memory. 30-minute setup.	Founders
MoizIbnYousaf/marketing-cli	25	51 skills, 5 research agents, brand memory that compounds, /cmo CLI entry point, local Studio dashboard	CMO-level ops

Sources: github.com/topics/ai-marketing + Cody Schneider (Agent Native podcast)

Paid advertising is the area with the most mature open-source agent infrastructure. Two repos cover it directly, and the Andromeda context from Cody Schneider explains why the agent approach is now viable without the $10k/month agency.

AgriciDaniel/claude-ads: 250+ checks across 7 ad platforms

Repo: github.com/AgriciDaniel/claude-ads

Stars: 6,200 (most starred ai-marketing repo on GitHub)

License: open-source

What it does: comprehensive paid advertising audit and optimisation skill for Claude Code. 250+ checks across Google, Meta, YouTube, LinkedIn, TikTok, Microsoft, and Apple Ads. Weighted scoring, parallel agents, industry-specific templates, and AI creative generation.

This is the same skill included in the Marketing Cockpit (Littlpinguin) and referenced in the 60 Claude Code Agents lead mag under @market-analyst. At 6.2k stars it is the most validated advertising skill in the Claude ecosystem.

Install:

npx skills add AgriciDaniel/claude-ads --global

What the Andromeda context means for this skill: Facebook's Andromeda algorithm reads ad creative (image, text, video, script) and the landing page together to decide who sees the ad. You no longer need interest-based targeting. The agent researches pain points, generates on-brand static and video creative, publishes it via the Facebook Marketing API, turns off losers after a 2-3 day signal window, promotes winners, and feeds winning creative back into the generation loop as training data. The claude-ads skill handles the audit and optimisation layer of this.

Anil-matcha/Open-AI-UGC: open-source AI video ad generation

Repo: github.com/Anil-matcha/Open-AI-UGC

Stars: 117

Stack: Next.js, Stripe, MUAPI

License: open-source, self-host

What it does: free open-source alternative to Arcads and MakeUGC. Generates AI UGC video ads with realistic AI actors using Veo 3.1, Seedance 2, Grok Video, and Happy Horse 1.

This is the open-source version of the video pipeline Cody Schneider describes using HeyGen and Seedance for. The pain point he identifies: Seedance clips are under 9 seconds, so you have to stitch frames together for 30-second ads. Open-AI-UGC handles the stitching and the actor layer. You bring the pain-point research and the script from the claude-ads or ai-marketing-claude skill.

Self-host setup:

git clone https://github.com/Anil-matcha/Open-AI-UGC
cp .env.example .env
# Add MUAPI key, Stripe keys, video model API keys
npm install && npm run dev

GTM use case: run the pain-point research agent (reddit scraping + competitor ad analysis) to identify the top 3 pain points, then use Open-AI-UGC to generate 5-10 UGC video variants per pain point. Publish to Facebook via the Marketing API. Let the data warehouse determine which pain point angle converts.

The full Facebook ads agent architecture (Cody Schneider)

Cody's production setup for running a fully autonomous Facebook ads agent:

1. Pain point research: scrape Reddit for what the target customer is complaining about. Use Perplexity or Claude to rank the top 3 most-referenced pain points. These seed all creative.
2. Static creative generation: use a vision model (Kai AI, Google Nano Banana, or equivalent) to bulk-generate static ads. Feed it a brand style guide (fonts, colours, brand rules) and reference ads from competitors or adjacent industries. Run a vision model over outputs to QA: is the text readable, are fonts on-brand, does it match the style guide?
3. Video creative generation: AI avatar UGC scripts from the pain point research. HeyGen or Seedance for the video. Stitch clips together for 30-second ads.
4. Publish via Facebook Marketing API: use the API only for writes (publishing content, turning off ads, promoting winners). Never use the API to pull large volumes of data. That is what gets accounts banned.
5. Data warehouse loop: AirByte (data pipeline, open source, self-hosted) pulls from Facebook Ads, Google Analytics, PostHog, HubSpot, and Stripe into ClickHouse (data warehouse, open source, self-hosted). The agent reads from the warehouse, not directly from Facebook.
6. The learning loop: agent publishes 10 ads (2 adsets, 5 ads per adset), waits 2-3 days for signal, turns off worst performers, promotes winners. Winners go into a winners pool competing for available budget. Agent reads winning ad JSON prompts and scripts to understand what is working and generates more like them.
7. Solving for entropy: feed competitor ads from the Facebook Ads Library as new DNA. Pull YouTube and podcast transcripts from your category to surface new angles. Use ViralLow API to scrape the most viral TikTok and Reels content in your category from the last week.
8. Cloud hosting: deploy the agent on Railway, Heroku, or any cloud. This is just code with a decision loop. It does not have to run on a local Mac Mini.

Sources: github.com/topics/ai-marketing repos

Five repos covering LinkedIn growth, blog content with AI citation optimisation, social media captions, and the full-stack content pipeline.

sergebulaev/linkedin-skills: human-sounding LinkedIn posts from Claude Code

Repo: github.com/sergebulaev/linkedin-skills

Stars: 185

What it does: Claude Code skills specifically for LinkedIn growth. Writes human-sounding posts, crafts comments that get noticed, analyses your feed for what is performing, and builds a publishing cadence.

Target: content creators, founders, and marketers using Claude Code for LinkedIn. Plug-and-play skills for the terminal.

Install:

npx skills add sergebulaev/linkedin-skills --global

GTM use case: pairs with the Claude + Prosp lead mag. Use linkedin-skills to generate the content, Prosp to activate the warm audience the content builds.

AgriciDaniel/linkedin-content-creator: full LinkedIn content pipeline with direct posting

Repo: github.com/AgriciDaniel/linkedin-content-creator

Stars: 60

Stack: React, TypeScript, Gemini AI

What it does: text, image, and carousel LinkedIn post generation with AI research, a content calendar, batch generation, and direct LinkedIn posting via OAuth. More than a skill file, it is a full application.

GTM use case: batch generate 30 days of LinkedIn posts in one session. Schedule directly without leaving the tool.

AgriciDaniel/claude-blog: 30 sub-skills with 5-gate delivery contract

Repo: github.com/AgriciDaniel/claude-blog

Stars: 1,100

What it does: 30 sub-skills, 5 agents, and a 5-gate Blog Delivery Contract before any post publishes: validated brief, draft, quality scoring, AI detection check, fact verification. Dual-optimised for Google rankings and AI citations (ChatGPT, Perplexity, Claude, Gemini).

This is the same skill aggregated into the Marketing Cockpit (Littlpinguin). Already noted in that lead mag. Install:

npx skills add AgriciDaniel/claude-blog --global

GTM use case: for Conigma ICP companies running content-led outbound. Blog posts that rank on AI search engines are the top-of-funnel equivalent of organic LinkedIn reach. The 5-gate contract prevents AI slop from publishing.

rediumvex/social-media-caption-generator-claude: algorithm-optimised captions

Repo: github.com/rediumvex/social-media-caption-generator-claude

Stars: 73

Creator: @theromanknox (280K+ Instagram followers)

What it does: turns any text, script, or screenshot into ready-to-post captions for Instagram Reels, Carousel, TikTok, Threads, Facebook, and YouTube Shorts. Algorithm-optimised for shares, saves, and completion rate, not just engagement.

Why the creator matters: built by someone with 280K+ Instagram followers. The optimisation logic comes from real platform performance data, not generic best-practice advice.

GTM use case: takes the pain-point research from the ad agent and repurposes it into organic short-form captions. Same research, multiple distribution channels.

stevenflanagan1/social-ai-team: complete AI social team for SMBs

Repo: github.com/stevenflanagan1/social-ai-team

Stars: 136

What it does: a complete AI social team built as Claude Code skills for SMBs. Brand setup, content calendar, captions, creative, and performance review.

GTM use case: the lightest-weight complete social stack available. If you do not want to build a full cockpit (Littlpinguin) or run a 50-skill system, this is the 80/20 version.

Sources: github.com/topics/ai-marketing repos

Three repos for the research and intelligence layer: competitor analysis, lead generation lists, and market research.

ScaleBrick/founder-marketing-skills: keyword research and competitive analysis

Repo: github.com/ScaleBrick/founder-marketing-skills

Stars: 77

What it does: open-source marketing skills for founders. Keyword research, growth strategy, social search audit (TikTok and Instagram), and competitor analysis. Install with npx skills.

Install:

npx skills add ScaleBrick/founder-marketing-skills --global

GTM use case: pre-campaign research layer. Use these skills to identify the market landscape before building ad creative or content. The social search audit gives you TikTok and Instagram trend data to feed the entropy-solving loop Cody Schneider describes.

toofast1/awesome-ai-lead-generation: the modern growth stack curated

Repo: github.com/toofast1/awesome-ai-lead-generation

Stars: 48

What it does: a curated awesome-list of AI tools for finding customers, social listening, and automated outreach. Covers cold email, lead generation, B2B sales automation, and social listening tools.

GTM use case: use as a reference when building your outbound stack. Covers the tools layer above what Claude Code skills handle: the enrichment providers, sequencers, and data sources that feed the agents.

AgriciDaniel/claude-email: inbox triage to deliverability audit

Repo: github.com/AgriciDaniel/claude-email

Stars: 75

What it does: AI-powered email management and marketing skill for Claude Code. Inbox triage, email composition, quality review, deliverability audit, automation sequences, and marketing strategy.

Install:

npx skills add AgriciDaniel/claude-email --global

GTM use case: the email layer of the full outbound stack. Pairs with the Claude Code for B2B Cold Email lead mag (personalisation waterfall, Prospeo enrichment, sequencer push) and the Claude + Prosp lead mag (warm DM conversion).

Sources: github.com/topics/ai-marketing repos

Four repos that provide full marketing stacks or agent workspace infrastructure rather than individual skills.

zubair-trabzada/ai-marketing-claude: 15 skills with parallel subagents

Repo: github.com/zubair-trabzada/ai-marketing-claude

Stars: 1,900

What it does: 15 marketing skills with parallel subagents. Website audit, copy generation, email sequences, ad campaigns, content calendars, competitive intelligence, and client-ready PDF reports.

GTM distinction from the Marketing Cockpit (Littlpinguin): lighter weight, no brand doctrine gate, no PostToolUse enforcement hook. Better for individual GTM engineers who want skills without the full multi-brand agency infrastructure.

GTM use case: the fastest path from zero to a working marketing skill set. 15 skills covering the full stack, ready to use in one install.

indranilbanerjee/digital-marketing-pro: 158 skills, 25 agents, AEO/GEO optimised

Repo: github.com/indranilbanerjee/digital-marketing-pro

Stars: 145

License: MIT

What it does: open-source AI marketing plugin for agencies and in-house teams. 158 skills, 25 specialist agents, a 12-part strategy flow. 6-platform AEO/GEO including Google AI Mode. EU AI Act Article 50 compliant. Installs on Claude Code, Cowork, Codex, Cursor, Copilot CLI, and Antigravity.

Key differentiator: the AEO/GEO layer. Answer Engine Optimisation and Generative Engine Optimisation for AI search (ChatGPT, Perplexity, Claude, Google AI Mode) is the content moat for 2026. This is the only repo in the list that specifically builds for AI citation acquisition across all 6 major platforms.

GTM use case: for agencies running content for multiple clients who need the broadest skill coverage across all channels and platforms.

ericosiu/marketing-os-starter: AI marketing team in 30 minutes

Repo: github.com/ericosiu/marketing-os-starter

Stars: 42

What it does: a cloneable AI marketing team. 4 specialised agents, 5 skills, structured handoffs, persistent memory. Built on Claude Code. Setup in 30 minutes.

This is the same repo referenced in the 60 Claude Code Agents lead mag (ericosiu). The growth and sales depth it covers is different from the Marketing Cockpit: better for sales pipeline and outbound, lighter on content and brand.

GTM use case: founders who want a working multi-agent marketing setup quickly without building from scratch. The structured handoffs make it a good starting point to customise.

SaigonXIII/evc and MoizIbnYousaf/marketing-cli: agent workspaces

evc (SaigonXIII, 48 stars): open-source marketing agent workspace for Claude Code. 42 commands, 12 hooks, 4 industry starters. The hooks are the differentiator: similar to the Marketing Cockpit's PostToolUse gate, evc uses 12 hooks to enforce workflow behaviour automatically.

marketing-cli (MoizIbnYousaf, 25 stars): 51 skills, 5 research agents, brand memory that compounds, /cmo CLI entry point, local Studio dashboard in beta. TypeScript. The /cmo entry point makes it the most natural-language-friendly of the workspace options.

# evc
git clone https://github.com/SaigonXIII/evc

# marketing-cli
npx marketing-cli install
# Then invoke with:
/cmo

All repos, install commands, and how to choose what to install.

Which repo for which GTM job

Job to be done	Repo	Stars	Install
Paid ads audit and optimisation (all platforms)	AgriciDaniel/claude-ads	6.2k	npx skills add AgriciDaniel/claude-ads
AI UGC video ad generation	Anil-matcha/Open-AI-UGC	117	git clone + self-host
LinkedIn posts and comments from Claude Code	sergebulaev/linkedin-skills	185	npx skills add sergebulaev/linkedin-skills
LinkedIn content calendar + direct posting	AgriciDaniel/linkedin-content-creator	60	git clone
Blog content with 5-gate quality contract	AgriciDaniel/claude-blog	1.1k	npx skills add AgriciDaniel/claude-blog
Social captions optimised for algorithm	rediumvex/social-media-caption-generator-claude	73	npx skills add
Complete SMB social team	stevenflanagan1/social-ai-team	136	git clone
Keyword research and competitor analysis	ScaleBrick/founder-marketing-skills	77	npx skills add ScaleBrick/founder-marketing-skills
Curated outbound and lead gen tool list	toofast1/awesome-ai-lead-generation	48	browse / reference
Email triage, composition, deliverability	AgriciDaniel/claude-email	75	npx skills add AgriciDaniel/claude-email
15-skill full-stack marketing suite	zubair-trabzada/ai-marketing-claude	1.9k	git clone
158 skills, AEO/GEO, 6-platform AI search	indranilbanerjee/digital-marketing-pro	145	git clone
4-agent marketing team, 30 min setup	ericosiu/marketing-os-starter	42	git clone
Agent workspace with 42 commands and 12 hooks	SaigonXIII/evc	48	git clone
/cmo CLI with brand memory	MoizIbnYousaf/marketing-cli	25	npx marketing-cli install

The real marketing agent infrastructure (Cody Schneider)

The GitHub repos above handle the skill and agent layer. Below is the data infrastructure layer required to run autonomous agents that improve over time rather than just executing tasks.

# DATA PIPELINE: AirByte (open source, self-hosted)
# Connects Facebook Ads, Google Analytics, PostHog, HubSpot, Stripe
# into a single data warehouse
# github.com/airbytehq/airbyte

# DATA WAREHOUSE: ClickHouse (open source, self-hosted)
# The single source of truth all agents read from
# github.com/ClickHouse/ClickHouse

# AGENT HOSTING: Railway or Heroku
# Agents are just code with a decision loop
# Deploy like any other application
# No Mac Mini required

# FACEBOOK ADS API: write-only
# Use only for: publishing content, turning off ads, promoting winners
# NEVER use for bulk data reads (account ban risk)
# Read ad performance data from ClickHouse, not from the Facebook API directly

# ENTROPY SOLVING: three sources of new creative DNA
# 1. Facebook Ads Library: competitor ad scripts and formats
# 2. YouTube and podcast transcripts from your category
# 3. ViralLow API: most viral TikTok and Reels content from last week

# THE AD PUBLISHING CADENCE
# 2 adsets per day, 5 ads per adset = 10 new ads daily
# 2-3 day signal window before any optimisation decision
# Turn off worst performers, promote winners
# Winners pool competes for budget
# Agent analyses winning JSON prompts and scripts to learn what converts

What to build first

For a GTM engineer or B2B founder starting from scratch:

Week 1 (skills layer): install 3-4 skills that address your most immediate bottleneck. If content is the bottleneck: claude-blog + linkedin-skills + social-media-caption-generator. If outbound is the bottleneck: claude-ads + claude-email + founder-marketing-skills.

Week 2 (data layer): set up AirByte and ClickHouse. Connect your ad platform, analytics, and CRM. This is what turns a skill from a one-shot tool into an agent that improves over time.

Week 3 (agent layer): deploy the first autonomous agent in the cloud. The Facebook ads loop is the highest-ROI starting point if you are running paid. The content calendar loop is the highest-ROI starting point if you are running organic.

Week 4+ (entropy solving): set up the competitor ad library scrape and the ViralLow feed. Without these, the agent converges on the same creative after 2-3 weeks.

The three-part test for whether you have built a real marketing agent (Cody Schneider):

1. Does it have a data pipeline giving it unified visibility of the whole pipeline?
2. Is it making autonomous decisions on a cadence with a thinking loop?
3. Is it hosted in the cloud so it runs continuously and improves from feedback?

If the answer to all three is yes, you have built a marketing agent. If not, you have a skill or an automation.