This project automates video generation and uploading to multiple platforms:

• YouTube (Data API v3)
• Instagram Reels (Graph API)
• Chinese platforms: Xiaohongshu/小红书 (XHS), Douyin, Toutiao
• Optional AI generation for video topics and scripts

New: Native support for Xiaohongshu uploads via MCP (Model Context Protocol) using the official MCP SDK and the open-source xiaohongshu-mcp server.

Reference: https://github.com/xpzouying/xiaohongshu-mcp

• OAuth 2.0 for YouTube and Instagram
• Automated video generation pipeline (topic → script → voice → render)
• Upload to YouTube, Instagram, and Chinese platforms
• Xiaohongshu (XHS) via MCP SDK (fast, reliable, Claude-compatible transport)
• Environment-based configuration (.env.development, .env.production)
• Email notifications on successful uploads

• Python 3.10+ recommended
• macOS, Windows, or Linux

Install Python dependencies:

pip install -r requirements.txt

Install Playwright browsers if you plan to upload via non-MCP browser automation (optional):

python -m playwright install chromium

This project uses separate env files based on ENV:

• .env.development (default)
• .env.production

Set which file to load by exporting ENV:

# macOS/Linux
export ENV=production
# Windows (PowerShell)
$env:ENV = "production"

• OPEN_AI_KEY
• IG_USER_ID, IG_ACCESS_TOKEN (Instagram Upload)
• SMTP credentials if using email notifications

See example.env for guidance and copy values into .env.development or .env.production.

Add these to your env file or export them before running:

• XHS_MCP_ENABLED=true — enable MCP mode for XHS
• XHS_MCP_SERVER_URL=http://<mcp-host>:18060/mcp — MCP HTTP endpoint
• XHS_MCP_VIDEO_DIR=... — absolute directory path on the MCP server where videos are stored
• HEADLESS=false — optional; if false, non-MCP browser automation can show a browser window

Example (Windows MCP host):

set "XHS_MCP_ENABLED=true"
set "XHS_MCP_SERVER_URL=http://192.168.1.9:18060/mcp"
set "XHS_MCP_VIDEO_DIR=C:\\Users\\jack\\Python-Apps\\youtube-uploader\\downloaded_videos\\pending_upload"

Example (macOS/Linux shell):

export XHS_MCP_ENABLED=true
export XHS_MCP_SERVER_URL=http://192.168.1.9:18060/mcp
export XHS_MCP_VIDEO_DIR="/mnt/shared/pending_upload"  # path as seen by the MCP server

Important: XHS_MCP_VIDEO_DIR must be a path that exists on the MCP server host. The app downloads the generated video directly into that folder so the MCP server can access it instantly.

From the project root:

# English video: generate + YouTube + Instagram (depending on env flags)
python main.py --language en

# Chinese video: generate + upload to Chinese platforms (XHS via MCP if enabled)
python main.py --language zh

Windows helper for Chinese uploads (uses MCP by default and maps paths correctly):

• run_cn.bat sets the required env vars and runs python main.py --language zh.

This project integrates with the xiaohongshu-mcp server using the official MCP Python SDK.

How it works:

1. The app generates your video.
2. If MCP mode is enabled, the video is downloaded directly into XHS_MCP_VIDEO_DIR (a folder on the MCP server host).
3. The uploader calls the MCP tool publish_with_video with the server-side file path.
4. The MCP server automates the XHS web upload.

• Deploy the server: https://github.com/xpzouying/xiaohongshu-mcp
• Start it with headless=false during debugging to see the browser.
• Log in to Xiaohongshu once in that browser session.
• Confirm the server is reachable at http://<host>:18060/mcp.

Run the MCP test script (path must exist on the MCP server):

python test_mcp_upload.py "C:\\Users\\jack\\Python-Apps\\youtube-uploader\\downloaded_videos\\pending_upload\\example.mp4"

The script will:

• Connect to the MCP server
• Verify login (check_login_status)
• Attempt a publish using publish_with_video

• Browser not visible: ensure the MCP server was started with headless=false.
• "Upload button not present": start with headless=false and watch the automation; element locators may need updating on the MCP server side.
• Slow upload (much slower than manual): ensure the MCP host has good network and disk performance. Keep the video file on the same machine (we already do by writing into XHS_MCP_VIDEO_DIR). Disable VPN/Proxy and try during off-peak hours.
• Timeout: our client uses a 15-minute publish timeout. If your network is slow, increase it in platforms/xhs/uploader_mcp_final.py if needed.

We adjusted dependency versions to align the MCP SDK with httpx/anyio/pydantic:

• httpx: >=0.27.1,<0.28
• anyio: >=4.6,<5
• mcp SDK: >=1.16.0,<2.0
• pydantic: >=2.10.1,<3

Install:

pip install -r requirements.txt

If Playwright is used by your non-MCP uploaders, install the Chromium runtime:

python -m playwright install chromium

• main.py — orchestrates generation and uploads; maps paths for MCP if enabled
• main_cn.py — async upload entry for Chinese platforms
• platforms/xhs/uploader_mcp_final.py — MCP-based XHS uploader using official SDK
• test_mcp_upload.py — test harness for MCP upload
• run_cn.bat — Windows helper to run Chinese/MCP workflow
• downloaded_videos/ — local downloads (non-MCP) or intermediate files

• Do not commit .env.* files or secrets
• Use separate .env.development and .env.production
• Rotate API keys regularly

• Google YouTube Data API, Facebook Graph API
• Chinese platform uploader references from https://github.com/aceliuchanghong/bili_douyin_xhs_uploader
• Xiaohongshu MCP server: https://github.com/xpzouying/xiaohongshu-mcp

MIT. See LICENSE.