MarketMeNow

The marketing intern that never sleeps.

Open-source framework that generates and publishes marketing content across 7 platforms from a single command.

---

Marketing was eating 2+ hours/day as a solo founder. So I built an intern that never sleeps.

Click to watch an AI-generated Instagram Reel

Platform	Content	Generate	Publish	In-Context Learning
Instagram	Reels, Carousels	✅	✅	✅
X / Twitter	Replies, Threads	✅	✅	✅
Reddit	Comments, Posts	✅	✅	-
LinkedIn	Posts, Images, Videos, Docs	✅	🚧 WIP	-
YouTube	Shorts	✅	✅	-
TikTok	Videos	✅	✅	-
Email	Bulk outreach	✅	✅	-

Features

In-Context Learning Learns from your top-performing posts. The more you post, the better it matches your voice.	Brand Templates Figma MCP + YAML templates. Your fonts, colors, layout. AI fills the content.	Engagement Automation Finds relevant conversations, writes contextual replies, posts with human-like timing.
Email Batching CSV in, 100 personalized emails out. Picks up where it left off.	Real-Time Dashboard Live progress, streaming logs, approve/reject per item. One click to publish all.	Modular Adapters Add a platform with zero changes to core. Ports-and-adapters architecture.
Auto-Heal One command runs lint, format, and tests - then hands failures to the Cursor agent to fix automatically.	Content Capsules Every generated reel, carousel, and thread is packaged into a self-contained capsule. Cross-post to any platform with one command — no re-generation needed.	Cross-Platform Repurposing Repurpose any capsule for a different platform and format. Video caption becomes a Twitter thread. Thread becomes a LinkedIn post. LLM-powered format-aware transformations.

Quick Start

git clone https://github.com/thearnavrustagi/marketmenow.git && cd marketmenow && bash setup.sh

Edit .env with your API keys, then:

uv run mmn-web          # Dashboard at http://localhost:8000

Requirements: Python 3.12+, uv (setup.sh installs it). Docker recommended for PostgreSQL. Node.js 18+ only if you want Instagram Reels.

Manual setup

git clone https://github.com/thearnavrustagi/marketmenow.git
cd marketmenow

uv sync
docker compose up -d
cp .env.example .env

uv run pre-commit install --hook-type pre-push   # tests + lint before every push
uv run playwright install chromium
cd src/adapters/instagram/reels/remotion && npm install && cd -

uv run mmn-web

Platform credentials

You only need credentials for the platforms you use:

Platform	What you need
Instagram	INSTAGRAM_ACCESS_TOKEN, INSTAGRAM_BUSINESS_ACCOUNT_ID
Facebook	FACEBOOK_C_USER, FACEBOOK_XS (optionally FACEBOOK_PAGE_IDS)
Twitter/X	TWITTER_AUTH_TOKEN, TWITTER_CT0 (or mmn twitter login)
Reddit	REDDIT_SESSION cookie, REDDIT_USERNAME
LinkedIn	LINKEDIN_ACCESS_TOKEN (or LINKEDIN_LI_AT cookie)
YouTube	Google OAuth 2.0 (mmn auth youtube)
TikTok	TIKTOK_SESSION_ID cookie (or OAuth 2.0 via mmn auth tiktok)
Email	SMTP_HOST, SMTP_PORT, SMTP_USERNAME, SMTP_PASSWORD, SMTP_FROM
AI (all)	AI Studio: GEMINI_API_KEY (or GOOGLE_API_KEY) or Vertex: GOOGLE_APPLICATION_CREDENTIALS, VERTEX_AI_PROJECT

CLI reference

# Instagram
mmn reel create --publish
mmn carousel generate --publish

# Facebook
mmn facebook login --cookies
mmn facebook page-post --page your-page-slug --text "Hello!"

# Twitter/X
mmn twitter login
mmn twitter all
mmn twitter engage
mmn twitter reply -f replies.csv
mmn twitter thread --post

# Reddit
mmn reddit engage
mmn reddit reply -f comments.csv

# LinkedIn
mmn linkedin auth
mmn linkedin post --text "Hello!"

# YouTube
mmn youtube auth
mmn youtube upload video.mp4

# TikTok
mmn tiktok login --cookies         # inject sessionid from .env
mmn tiktok login                   # manual browser login
mmn tiktok auth                    # OAuth 2.0 (requires developer app)
mmn tiktok upload video.mp4 --title "caption" --hashtags "fyp,viral"

# Email
mmn email send -f contacts.csv -t template.html -r 0-100

# Content Capsules — repost any content to any platform
mmn capsule list                                          # list all capsules
mmn capsule info <capsule-id>                             # show capsule details
mmn run post-capsule --capsule <id> --platform youtube    # cross-post to YouTube
mmn run post-capsule --capsule <id> --platform tiktok     # cross-post to TikTok

# Cross-Platform Repurposing — adapt content format for different platforms
mmn run repurpose-capsule --capsule <id> --platform twitter     # video/post → thread
mmn run repurpose-capsule --capsule <id> --platform linkedin    # thread/video → text post
mmn run repurpose-capsule --capsule <id> --platform reddit      # any → text post
mmn run repurpose-capsule --capsule <id> --platform twitter --target-modality text_post  # override format

# Auto-heal (lint + format + test, then auto-fix via Cursor agent)
mmn heal              # fix issues automatically
mmn heal --no-fix     # report only, don't invoke the agent
mmn heal --verbose    # show full lint/test output

Customizing for your brand

All brand identity lives in YAML files, not code. Each product gets its own directory under projects/ with prompts, targets, personas, and templates that every workflow picks up automatically.

The easy way - interactive onboarding

mmn project add my-product

A 10-phase wizard walks you through everything:

1. Brand identity - name, URL, tagline, colors
2. Product features - one per line
3. Target customer - ICP, pain points, keywords, platforms
4. Social persona - voice, tone, example phrases
5. Twitter targets - accounts and hashtags to engage
6. Reddit targets - subreddits and keywords
7. Outreach profile - rubric, message style, rate limits
8. Platform prompts - auto-generated from your brand + persona via AI
9. Reel templates - auto-generated video structure + companion prompt
10. Summary - review, activate, done

Switch between products any time with mmn project use <slug>. Workflows, prompts, and templates resolve from the active project first, falling back to globals.

The hard way - manual YAML editing

Prompts (text content): Prompts use a decomposed architecture: a persona YAML (personas/default.yaml or prompts/<platform>/persona.yaml) defines voice, tone, and example phrases, while function templates (prompts/<platform>/reply_generation.yaml, comment_generation.yaml, script_generation.yaml, etc.) define the task-specific instructions. Edit these in projects/<slug>/ to customize per product. Or use the meta-prompt in prompts/prompt.md to generate entire prompt YAMLs with any AI chat.

Reels (video content): Use the reel template meta-prompt in src/adapters/instagram/reels/templates/prompt.md to generate a template YAML (scenes, beats, transitions) and a companion prompt YAML (the AI persona). Drop them into src/adapters/instagram/reels/templates/ and prompts/instagram/, then mmn reel create --template your_id.

Architecture

Ports-and-adapters design. The core engine knows nothing about any specific platform. Each adapter implements PlatformAdapter, ContentRenderer, and Uploader protocols.

graph LR
    subgraph core [Core]
        direction TB
        Scheduler --> Orchestrator
        Orchestrator --> Pipeline
        Pipeline --> Normaliser
    end

    subgraph ports [Protocols]
        direction TB
        Renderer["ContentRenderer"]
        Sanitiser["TextSanitiser"]
        Uploader["Uploader"]
        Adapter["PlatformAdapter"]
    end

    subgraph adapters [Adapters]
        direction TB
        IG["Instagram"]
        TW["Twitter / X"]
        RD["Reddit"]
        LI["LinkedIn"]
        YT["YouTube"]
        TT["TikTok"]
        EM["Email / SMTP"]
    end

    Pipeline --> Renderer
    Renderer --> Sanitiser
    Sanitiser --> Uploader
    Uploader --> Adapter
    IG -.-> Adapter
    TW -.-> Adapter
    RD -.-> Adapter
    LI -.-> Adapter
    YT -.-> Adapter
    TT -.-> Adapter
    EM -.-> Adapter

Pipeline: Normalise → Render → Sanitise → Upload → Publish

The Sanitise step strips em-dashes, en-dashes, and other AI-telltale formatting from all text fields before upload - a simple anti-AI-detection layer that runs automatically on every piece of content.

Adding a platform: Create src/adapters/yourplatform/, implement the protocols, register with AdapterRegistry. Zero changes to core.

Roadmap

Checked items are shipped. Unchecked items are planned or in progress.

Done

• 7-platform content engine, in-context learning, ports-and-adapters core, personas, Twitter discovery & cold DM, composable pipelines, auto-heal, Reddit & LinkedIn publish
• Exploration/Exploitation - epsilon-greedy ICL with farthest-point embedding sampling for diverse example selection, per-reply explore/exploit toggle, performance tracking

Up Next

• Unified PromptBuilder interface - migrate all adapters to the composable PromptBuilder (persona + function + ICL blocks), replacing legacy direct YAML prompt loading across Instagram Reels, Reddit, Facebook, LinkedIn, email, and outreach
• Extend ICL to all platforms - bring epsilon-greedy in-context learning (currently Twitter-only) to Instagram, Reddit, LinkedIn, and other adapters
• A/B testing - generate content variants, publish them across splits, and measure which performs better to continuously optimise messaging and format
• Cross-platform content repurposing - adapt a single piece of content across platforms automatically (e.g. LinkedIn post → Twitter thread → Instagram carousel) with LLM-powered format-aware transformations via the repurpose-capsule workflow
• Analytics dashboard - unified cross-platform analytics view to track performance, compare content types, and surface actionable insights across all platforms
• Analytics feedback loop - auto-collect post performance metrics and feed them back to rank ICL examples, so the system improves with every publish
• Full Autonomous Campaign - user types "Run a 30-day lead-gen campaign for my AI tool" and the agent does everything: researches audience, plans 12+ posts + repurposing chain, generates capsules, schedules, monitors, optimizes, sends weekly report. Record a 90-second demo video of end-to-end autonomous marketing
• Smart scheduling + best-time posting - auto-schedule capsules at optimal posting times based on historical engagement data. Add "Smart Schedule" button in dashboard
• Content calendar - visual calendar view showing planned, scheduled, and published content across all platforms. Drag-and-drop rescheduling
• Dashboard reel template builder - visual UI in the web dashboard to create and edit reel templates (scenes, beats, transitions) without touching YAML
• An evolving system - the system evolves with code repositories automatically

Contributing

See CONTRIBUTING.md for development setup, code style, and the PR process.

License

MIT