Content Factory Mcp

publish-mcp

AI-powered content factory as an MCP server. Generate, translate, and publish content to Telegram, VK, MAX, and Threads from Claude Code, Cursor, or any MCP client.

One config snippet. Full content pipeline: brand setup, competitor analysis, source discovery, AI writing, image generation, multi-channel publishing.

Quick Start

{
  "mcpServers": {
    "publish": {
      "command": "npx",
      "args": ["-y", "content-factory-mcp"],
      "env": {
        "TELEGRAM_BOT_TOKEN": "123456:ABC-DEF...",
        "TELEGRAM_CHAT_ID": "@my_channel",
        "GEMINI_API_KEY": "AIza..."
      }
    }
  }
}

Then ask your AI agent:

"Set up my brand for the AI education niche, find content sources, generate a post, create an image, and publish to Telegram"

Tools (11 total)

Content Factory

Tool	Description
setup_brand	Configure niche, tone of voice, audience, post structure, sample posts
get_brand	Show current brand profile
analyze_competitors	Scrape competitor pages, analyze their content strategy with AI
find_sources	Discover RSS feeds, scrape websites, or AI-discover sources for your niche
generate_post	Generate a post from topic or source material, using brand voice
translate_post	Translate posts to any language, maintaining brand voice
generate_media	Generate images via Gemini Imagen or DALL-E

Publishing

Tool	Description
publish	Publish text + optional image to a single channel
publish_all	Publish to all configured channels at once
list_channels	Show configured channels
preview	Dry-run: see what would be published

Full Pipeline Example

You (to Claude Code):
  "I run a fitness blog. Set up my brand, analyze my competitor
   @fitnessguru, find content sources, write a post about
   morning routines, generate an image, and publish to Telegram."

Claude Code calls:
  1. setup_brand(niche="fitness", tone="motivational", ...)
  2. analyze_competitors(urls=["https://fitnessguru.com"])
  3. find_sources(action="discover", niche="fitness")
  4. generate_post(topic="5 morning routines for energy")
  5. generate_media(post_text="<generated post>")
  6. publish(text="<post>", image_url="<image>", channel="telegram")

Environment Variables

Publishing (at least one platform required)

TELEGRAM_BOT_TOKEN=123456:ABC-DEF...
TELEGRAM_CHAT_ID=@my_channel

VK_ACCESS_TOKEN=vk1.a.xxx
VK_GROUP_ID=12345678
VK_USER_TOKEN=vk1.a.yyy          # optional, for photo uploads

MAX_BOT_TOKEN=xxx
MAX_CHAT_ID=-12345678

THREADS_ACCESS_TOKEN=xxx
THREADS_USER_ID=12345678

AI Backend (required for content factory tools)

Set ONE of these. Priority: explicit AI_PROVIDER > first key found.

# Gemini (free tier available)
GEMINI_API_KEY=AIza...
GEMINI_MODEL=gemini-2.5-flash         # optional

# OpenAI
OPENAI_API_KEY=sk-...
OPENAI_MODEL=gpt-4o-mini              # optional

# Claude (Anthropic)
ANTHROPIC_API_KEY=sk-ant-...
ANTHROPIC_MODEL=claude-sonnet-4-20250514  # optional

# Ollama (local, free, no API key)
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=llama3.1                  # optional

# Any OpenAI-compatible API (Together, Groq, Fireworks, LM Studio, etc.)
OPENAI_COMPATIBLE_BASE_URL=https://api.together.xyz/v1
OPENAI_COMPATIBLE_API_KEY=xxx
OPENAI_COMPATIBLE_MODEL=meta-llama/Llama-3.1-70B-Instruct

# Force specific provider (overrides auto-detection)
AI_PROVIDER=gemini|openai|anthropic|ollama|openai-compatible

Multi-Channel Config

PUBLISH_CHANNELS=[{"name":"my-tg","platform":"telegram","token":"...","chatId":"@chan"},{"name":"my-vk","platform":"vk","token":"...","chatId":"123","groupId":"123"}]

Storage

PUBLISH_MCP_DATA_DIR=/path/to/data  # optional, default: ./.publish-mcp

Platforms

Platform	Auth	Publishing
Telegram	Bot token + admin in channel	sendMessage / sendPhoto
VK	Community token (+ user token for photos)	wall.post
MAX	Bot token	sendMessage + photo
Threads	Long-lived access token	Container + publish API

Development

npm install
npm run build
node dist/index.js  # starts MCP server on stdio

License

MIT