meta-nexus
README.md
# ๐ฌ Meta Nexus Agent โ Zero-API Facebook & Instagram Autonomous Studio
<p align="center">
<a href="https://github.com/muhdfaisalwork-gif/meta-agent/actions"><img src="https://img.shields.io/badge/Tests-10%2F10%20Passing-brightgreen?logo=github-actions&logoColor=white" alt="Tests"></a>
<a href="https://modelcontextprotocol.io/"><img src="https://img.shields.io/badge/MCP-Protocol%202024--11--05-blue?logo=anthropic&logoColor=white" alt="MCP Server"></a>
<a href="https://ollama.com/"><img src="https://img.shields.io/badge/Ollama-100%25%20Offline%20Ready-black?logo=ollama&logoColor=white" alt="Ollama Local"></a>
<a href="https://github.com/yt-dlp/yt-dlp"><img src="https://img.shields.io/badge/yt--dlp-Saved%20Reels%20Extractor-red?logo=youtube&logoColor=white" alt="yt-dlp"></a>
<a href="https://ffmpeg.org/"><img src="https://img.shields.io/badge/FFmpeg-Audio%20%26%20Reel%20Studio-007808?logo=ffmpeg&logoColor=white" alt="FFmpeg"></a>
<a href="https://playwright.dev/"><img src="https://img.shields.io/badge/Playwright-Stealth%20Persistent-2EAD33?logo=playwright&logoColor=white" alt="Playwright"></a>
<a href="#license"><img src="https://img.shields.io/badge/License-MIT-blue.svg" alt="MIT License"></a>
</p>
<p align="center">
<b>The #1 Free, Open-Source Zero-API Alternative to Hootsuite, Later & ManyChat for Facebook & Instagram.</b><br>
Powered by <b>Playwright Persistent Chrome Context</b>, <b>yt-dlp</b>, <b>FFmpeg Audio Studio</b>, an <b>Automated Excel Master Sheet (`saved_videos_log.xlsx`)</b>, an <b>82-Rule Humanizer Gate</b>, <b>Local Ollama</b>, and native <b>Claude / Cursor MCP Server</b> support.
</p>
---
## โก Why Meta Nexus Agent?
Paid social media schedulers and comment automation tools charge **$49 to $199/month**, force you through brittle Meta Graph API approvals, strip trending audio tracks, and generate detectable AI slop that destroys Reel retention.
**Meta Nexus Agent solves this 100% locally with zero API fees:**
1. **Persistent Login Engine (`chrome_profile`)**: Log into Instagram and Facebook manually once (or inject `sessionid` / `c_user`+`xs` cookies in `< 1s`). Exports decrypted Netscape `cookies.txt` so `yt-dlp` never hits Windows Chrome DPAPI locks.
2. **Saved Video Extractor & `yt-dlp` Downloader (`02_saved_extractor.py`)**: Scrapes your Instagram Saved (`/saved/all-posts/`) and Facebook Saved (`/saved`) collections, downloads high-resolution `.mp4` Reels via `yt-dlp`, and extracts full metadata (Title/Hook, Caption, Creator Handle, Duration, Audio Track).
3. **Automated Master Sheet (`saved_videos_log.xlsx`)**: Bidirectionally syncs every scraped, mixed, and published Reel between a styled Excel workbook and local SQLite (`storage/meta_agent.db`).
4. **FFmpeg Music & Reel Studio (`03_media_music_mixer.py`)**:
- **Scenario A (Reel + Local MP3 Pre-Mix)**: Lowers original video audio to `20%` and overlays `.mp3` tracks from `music_library/` at `85%` volume.
- **Scenario B (Static Image + Music -> 15s Reel)**: Loops any `.jpg/.png` or auto-generated `1080x1920` (`9:16`) builder quote card with an `.mp3` track into a 15-second vertical Reel.
- **Scenario C (Native Platform Audio)**: Flags rows with `NATIVE_UI:<song_name>` so the Playwright uploader selects trending music directly inside Instagram/Facebook's native "Add Music" UI modal.
5. **82-Rule Humanizer & Cognitive Brain**: Eliminates all 82 AI buzzwords (*delve, tapestry, robust, pivotal, game-changer, landscape*) and grounds captions and comment replies in verified project receipts from your Story Bank.
6. **Autonomous IG & FB Publishers + AI Comment Engine (`04_ig` & `05_fb`)**: Publishes Reels/Posts and runs **Mode 1 (Inbox/Comment Auto-Reply)** and **Mode 2 (Niche Hashtag Outreach)** with `3.5sโ9.2s` human jitter, `5โ8s` Reel watch simulation, and `50โ90ms` per-character keystroke typing.
---
## ๐ Comparison: Meta Nexus Agent vs. Paid SaaS
| Feature | **Meta Nexus Agent** | **Hootsuite / Buffer** | **Later** | **ManyChat** |
| :--- | :---: | :---: | :---: | :---: |
| **Price** | **$0 / month (Free & Open Source)** | $99 โ $249 / mo | $25 โ $80 / mo | $15 โ $145 / mo |
| **Meta Graph API Required?** | **โ Zero-API (Persistent Browser + Cookie)** | โ Requires Business API | โ Requires API | โ Requires Webhooks |
| **Saved Reels `yt-dlp` Downloader** | **โ Built-in (`chrome_profile/cookies.txt`)** | โ No | โ No | โ No |
| **FFmpeg 20%/85% Audio Mixer** | **โ Built-in (Scenarios A, B & C)** | โ No | โ No | โ No |
| **Automated Excel Master Sheet** | **โ Live `saved_videos_log.xlsx` Sync** | โ CSV export only | โ No | โ No |
| **Local Offline AI (Ollama)** | **โ Yes (`qwen2.5` / `llama3.2` / `deepseek-r1`)** | โ Cloud only | โ Cloud only | โ Cloud only |
| **82-Rule Anti-Slop Humanizer** | **โ Strict 4-Pass Programmatic Scrub** | โ Generic AI | โ Generic AI | โ Flow templates |
| **Claude Desktop / Cursor MCP** | **โ Native JSON-RPC 2.0 Server** | โ No | โ No | โ No |
---
## ๐๏ธ System Architecture
```mermaid
flowchart TD
subgraph Interfaces ["Multi-Channel Interfaces"]
UI["FastAPI Web Studio (:8005)"]
CLI["Master CLI (main_agent.py)"]
MCP["Claude Desktop & Cursor (MCP Server)"]
Excel["Master Excel Sheet (saved_videos_log.xlsx)"]
end
subgraph Pipeline ["6-Phase Autonomous Meta Pipeline"]
P1["01_login_setup.py\n(Persistent Chrome + cookies.txt)"]
P2["02_saved_extractor.py\n(IG/FB Saved Scraper + yt-dlp)"]
P3["03_media_music_mixer.py\n(FFmpeg 20%/85% Mix + 9:16 Card Reel)"]
P4["04_ig_publisher_commenter.py\n(Instagram Reels & AI Commenter)"]
P5["05_fb_publisher_commenter.py\n(Facebook Reels & AI Commenter)"]
P6["Anti-Ban Guardrails\n(3.5-9.2s Jitter + 50-90ms Typing)"]
end
subgraph Intelligence ["Cognitive Brain & Multi-Model LLM"]
Brain["Story Bank Receipts + Voice Profile"]
Humanizer["82-Rule Strict Humanizer Gate"]
LLM["Local Ollama (qwen2.5 / llama3.2)\n+ OpenRouter Free Fallback"]
end
Interfaces --> Pipeline
Pipeline <--> Excel
Pipeline --> Intelligence
```
---
## ๐ Directory & File Architecture
```text
G:\Meta agent\
โ
โโโ chrome_profile\ # Persistent Chrome user data (FB & IG cookies + Netscape cookies.txt for yt-dlp)
โโโ saved_videos\ # Downloaded raw .mp4 files from FB & IG Saved collections
โ โโโ instagram\
โ โโโ facebook\
โโโ music_library\ # Local .mp3 songs/audio tracks to attach to Posts & Reels
โโโ ready_to_post\ # Final processed 1080x1920 Reels & Image-Music videos ready for upload
โโโ logs\ # Error logs and execution history
โโโ storage\ # Local SQLite database (meta_agent.db) synced with Excel
โ
โโโ saved_videos_log.xlsx # Master spreadsheet (ID, Platform, Hook, Caption, Creator, URL, MP4, Music, Status, Posted URL)
โโโ config.json # Usernames, posting schedules, human-delay timings, Ollama model name
โ
โโโ 01_login_setup.py # Phase 1: Run ONCE to log into Facebook & Instagram manually (or inject cookies)
โโโ 02_saved_extractor.py # Phase 2: Scrapes Saved posts, downloads videos via yt-dlp, updates Excel sheet
โโโ 03_media_music_mixer.py # Phase 3: FFmpeg engine to combine music + video or static image + music
โโโ 04_ig_publisher_commenter.py # Phase 4 & 5: Playwright UI automation for Instagram Reels, Posts & Comments
โโโ 05_fb_publisher_commenter.py # Phase 4 & 5: Playwright UI automation for Facebook Reels, Posts & Comments
โโโ main_agent.py # Master CLI / AI controller that orchestrates all modules
โโโ run.py # FastAPI Web Studio Dashboard launcher (:8005)
โโโ run_mcp.py # Model Context Protocol (MCP) Server for Claude Desktop & Cursor
โโโ launch.bat # 1-Click Windows Studio Launcher
```
---
## ๐ Quick Start (Up in < 2 Minutes)
### Option A: 1-Click Launch (Windows)
Double-click `launch.bat` in the repository root:
```cmd
launch.bat
```
### Option B: Terminal / Python (All Platforms)
```bash
# Clone the repository
git clone https://github.com/muhdfaisalwork-gif/meta-agent.git
cd meta-agent
# Install dependencies
pip install -r requirements.txt
playwright install chromium
# Launch Web Studio Server (http://localhost:8005)
python run.py
```
---
## ๐ ๏ธ Running the 6 Blueprint Modules via CLI
```bash
# Phase 1: Log into Instagram (Tab 1) & Facebook (Tab 2) once in persistent Chrome
python 01_login_setup.py
# Phase 2: Scrape Saved Reels/Videos via yt-dlp & update saved_videos_log.xlsx
python 02_saved_extractor.py --platform both --max 5
# Phase 3 (Scenario A): Batch mix downloaded Reels with local MP3s (20% original / 85% music)
python 03_media_music_mixer.py --scenario batch
# Phase 3 (Scenario B): Turn a 9:16 quote card / static image + MP3 into a 15s 1080x1920 Reel
python 03_media_music_mixer.py --scenario b --hook-text "How we cut order execution latency from 1.2s to 420ms"
# Phase 4 & 5: Publish Reels & Auto-Reply to Comments on Instagram and Facebook
python 04_ig_publisher_commenter.py --action publish
python 04_ig_publisher_commenter.py --action auto-reply
python 05_fb_publisher_commenter.py --action publish
python 05_fb_publisher_commenter.py --action auto-reply
# Run Full End-to-End Autonomous Cycle
python main_agent.py --auto
```
---
## ๐ Connect to Claude Desktop or Cursor (MCP)
Meta Nexus Agent includes a certified **Model Context Protocol (MCP 2024-11-05)** server:
```bash
python scripts/install_mcp.py
```
Or add manually to `claude_desktop_config.json`:
```json
{
"mcpServers": {
"meta-nexus": {
"command": "python",
"args": ["G:\\Meta agent\\run_mcp.py"]
}
}
}
```
---
## ๐งช Automated Test Suite (100% Pass Rate)
```bash
python -m pytest -v
```
---
## ๐ค Contributing & Security
Contributions are warmly welcomed! Please read [CONTRIBUTING.md](CONTRIBUTING.md) and review [SECURITY.md](SECURITY.md).
---
## ๐ License
Distributed under the MIT License. See [LICENSE](LICENSE) for full details.
This server cannot be deployed
Maintenance
ActivityMaintained
ResponsivenessNo issues