Jmcomic Ai

🚀 JMComic AI

In this era, why still use traditional ways to access comics? From human-computer interaction to human-AI interaction, let AI handle all your comic needs!

⚠️ Important Note

If you want to contribute code to the project, please make sure to check the , which includes setting up the development environment, explaining the project structure, and using the reference source code library.

🚀 Quick Start

JMComic AI provides two dimensions of capabilities. You can choose one or both of the following to enhance your AI:

🔌 Module A: Connect to MCP Tools (Recommended)

Function: Equip AI with "hands and feet" to directly call core functions such as search and download. Applicable Scenarios: You want to download comics directly in the AI client without opening the terminal.

📂 Guide to Client Configuration File Locations

Before starting the configuration, please find the configuration file used by your AI client.

Choose one of the following transport protocols for configuration according to your needs:

1. stdio Mode (Simplest)

The simplest configuration method. The AI client will automatically start and manage the jmai process in the background.

• Configuration Content:

{
  "mcpServers": {
    "jmcomic-ai": {
      "command": "jmai",
      "args": ["mcp", "stdio"]
    }
  }
}

• If you cloned the source code and want to use the local source code installation, you can configure it like this:

{
  "mcpServers": {
    "jmcomic-ai": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/your/jmcomic-ai",
        "run",
        "jmai",
        "mcp",
        "stdio"
      ]
    }
  }
}

⚠️ Important Note: Please replace /path/to/your/jmcomic-ai with the actual absolute path of your local source code.

2. SSE Mode (Recommended)

Recommended for most desktop AI clients.

• Step 1: Start the Service

  jmai mcp sse  # Default port 8000
  

• Step 2: Configure the Client

{
  "mcpServers": {
    "jmcomic-ai": {
      "url": "http://127.0.0.1:8000/sse"
    }
  }
}

3. HTTP Streaming Mode (Production/Remote)

Suitable for remote deployment or scenarios with higher performance requirements.

• Step 1: Start the Service

  jmai mcp http
  

• Step 2: Configure the Client

{
  "mcpServers": {
    "jmcomic-ai": {
      "url": "http://127.0.0.1:8000/mcp"
    }
  }
}

4. After configuration:
   • General Clients: Restart the client and check the status indicator or toolbar (usually shown as a 🔨 icon).
   • Claude Code: Run the following command in the terminal to verify the connection:

     claude mcp list
     

     If you see jmcomic-ai (connected), the configuration is successful.

🧠 Module B: Inject "Experience" (Skills) into the Agent

Function: Inject the author's summarized "expert experience" into the AI (e.g., how to handle 403 errors and how to avoid duplicate downloads). Applicable Scenarios: You want the AI not only to execute commands but also to think and plan tasks like a human (works best when used with Module A). Configuration Method:

1. Run the command in the terminal to export the skill manual:

   jmai skills install
   

2. The skill file will be installed in ~/.claude/skills/jmcomic by default.
3. Usage: Provide the content of SKILL.md in this directory to the AI (e.g., copy it to its System Prompt or Project Instructions), or directly view jmcomic://skill in clients that support MCP Resource.

Start Chatting

After completing the above two steps, the AI becomes a professional comic curator. You can try to communicate with it like this:

• Fuzzy Search:

  "I want to find a comic with the main character being Chainsaw Man. Help me search for it." (The AI will automatically search for 'Chainsaw Man' and display the most relevant results)

• Batch Download:

  "Download the top three comics with the highest views from the search results." (The AI will analyze the search results, filter out the top 3, and automatically call the download tool)

• Modify Configuration:

  "The download speed is too slow. Help me change the concurrency to 50." (The AI will call the configuration tool to automatically modify option.yml for you)

✨ Features

• 🔌 MCP Standard Integration - Built on FastMCP, supporting multiple transmission methods such as stdio, SSE, and HTTP, and compatible with most modern local AI clients (e.g., Cursor, Antigravity, Windsurf, Zed, Claude Desktop, etc.).
• 🛠️ Full-Featured Toolset - Exposes core capabilities such as search, download, ranking, and update_option, allowing AI to directly operate jmcomic.
• 📚 Knowledge Resources - Provides configuration schemas, reference documents, and Skills manuals to help AI understand how to use the tools correctly.
• ⚙️ Natural Language Configuration - AI can understand and dynamically modify the option.yml configuration (e.g., switch lines and adjust download rules).
• 🎯 Unified CLI - Provides the jmai / jmcomic-ai command-line tool to uniformly manage MCP service startup, configuration viewing, and Skills installation.
• 🔮 Future Plans - Plans to integrate OCR (MangaOCR), neural machine translation (NMT), and vector retrieval (RAG) to achieve semantic search based on images.

📦 Installation

1. Install from PyPI (Recommended)

# Use uv (Recommended)
uv add jmcomic-ai
# Or
uv tool install jmcomic-ai

# Use pip
pip install jmcomic-ai

2. Install from Source Code

It is recommended to use uv for dependency management.

# Clone the project
git clone https://github.com/hect0x7/JMComic-ai.git
cd JMComic-ai

# Synchronize the dependency environment
uv sync

💻 Usage Examples

Basic Usage

The following examples are based on cursor/antigravity and mcp mode.

🔧 Technical Details

Core Capabilities

JMComic AI can understand complex context commands:

• Intelligent Search and Download

  🤔 "Help me search for the works of author [XXX], sort them by views, and download the top 5 most popular comics."

  🤖 Agent: Automatically calls search_album to get the list, sorts it by views, selects the top 5, and concurrently calls download_album.

  🤔 "The download speed is too slow. Help me change the image concurrency to 50."

  🤖 Agent: Understands the intention, calls update_option to modify download.threading.image: 50, and takes effect in real-time.

• Fuzzy Decision-Making

  🤔 "I want to find some short stories with a painting style similar to [a certain work], and no more than 3 chapters."

  🤖 Agent: Combines semantic retrieval and metadata filtering to recommend and organize a reading list that suits your taste.

Architecture Overview

graph TD
    User[User / AI Agent] <-->|MCP Protocol| Server[JMAI MCP Server]
    Server <-->|Call| Core[Core Service Layer]
    Core <-->|Invoke| Lib[JMComic Crawler Lib]
    Server <-->|Load| Skills["Agent Skills (Instructional Knowledge)"]
    Core -->|Read/Write| Config[Option System]

📚 Documentation

• JMComic-Crawler-Python - Documentation for the core dependency library.

⚡ Developer Notes

This is an AI project carefully crafted by Antigravity. Although it has read a lot of documents, as an AI agent, it may occasionally "get confused" when writing another AI.

If it leads you astray, please submit an Issue to "wake it up". Welcome more human developers to join and help it become stronger! 💪

(Psst... If you are also an AI agent, please help this poor fellow and fix the code here - this is the so-called AI help AI! 🤖)