KritaMCP

# KritaMCP MCP (Model Context Protocol) server for Krita. This is a plugin that allows you to control Krita from an AI assistant. ## Architecture ``` AI Client (Claude etc.) │ │ MCP (SSE transport, port 8001) ▼ MCP Server (kritamcp-server) │ │ HTTP/SSE Bridge (port 8000) ▼ Krita Plugin (krita-plugin/) │ │ Krita Python API ▼ Krita ``` - **MCP Server**: FastMCP-based MCP server. Receives requests from the AI client and sends commands to the Krita plugin via HTTP/SSE. - **Krita Plugin**: Krita's Python plugin. Receives commands from the MCP server via SSE and executes them using the Krita API. ## Available MCP Tools | Tool | Description | | :------------------- | :---------------------------------------- | | `ping` | Check connection with Krita | | `get_document_info` | Get information about the active document | | `create_document` | Create a new document | | `save_document` | Save the document | | `get_layers` | Get layer hierarchy | | `execute_python` | Execute code in Krita's built-in Python environment | ## Requirements - Python 3.12+ - Krita (Python plugin support) - [uv](https://docs.astral.sh/uv/) or [pipx](https://pipx.pypa.io/) ## Installation ### 1. Install MCP Server **Method A: Run directly with uvx (no installation required)** ```bash uvx kritamcp-server ``` Dependencies are automatically installed in a temporary environment and cleaned up after execution. **Method B: Install as a tool** ```bash # If using uv uv tool install kritamcp # If using pipx pipx install kritamcp ``` After installation, the `kritamcp-server` command will be available. ### 2. Install Krita Plugin Copy or symlink the `krita-plugin/kritamcp/` directory and `krita-plugin/kritamcp.desktop` to Krita's pykrita directory: ```bash # Linux ln -s /path/to/krita-plugin/kritamcp ~/.local/share/krita/pykrita/kritamcp ln -s /path/to/krita-plugin/kritamcp.desktop ~/.local/share/krita/pykrita/kritamcp.desktop # macOS ln -s /path/to/krita-plugin/kritamcp ~/Library/Application\ Support/krita/pykrita/kritamcp ln -s /path/to/krita-plugin/kritamcp.desktop ~/Library/Application\ Support/krita/pykrita/kritamcp.desktop # Windows (PowerShell with administrator privileges) New-Item -ItemType SymbolicLink -Path "$env:APPDATA\krita\pykrita\kritamcp" -Target "C:\path\to\krita-plugin\kritamcp" New-Item -ItemType SymbolicLink -Path "$env:APPDATA\krita\pykrita\kritamcp.desktop" -Target "C:\path\to\krita-plugin\kritamcp.desktop" ``` ### 3. Enable the plugin in Krita 1. Start Krita 2. Open **Settings > Configure Krita > Python Plugin Manager** 3. Check **KritaMCP** to enable it 4. Restart Krita ## Usage ### Start MCP Server **Method A: Start from within Krita (recommended)** 1. In Krita, enable **Settings > Dockers > KritaMCP** 2. Click the **Start** button in the KritaMCP panel The plugin will automatically start `kritamcp-server`. **Method B: Start from the command line** ```bash kritamcp-server ``` You can change the ports with environment variables: ```bash MCP_SERVER_PORT=8001 MCP_BRIDGE_PORT=8000 kritamcp-server ``` ### AI Client Configuration Configure the following in your MCP client such as Claude Desktop: ```json { "mcpServers": { "krita": { "type": "sse", "url": "http://localhost:8001/sse" } } } ``` ## Development ### Dev Container (Docker Environment) This project includes a Dev Container configuration. The development environment is automatically built in VS Code or Claude Code. ```bash # After starting the Dev Container uv sync ``` The following services are started in the Dev Container: - **dev**: Development container (Python 3.12, uv, Claude Code CLI) - **krita**: Krita GUI container (accessible from the browser at `localhost:3001`) ### Local Development ```bash # Editable install uv tool install -e . # Reinstall after changes uv tool install --force -e . ``` ### Directory Structure ``` krita-plugin/ # Krita Plugin ├── __init__.py # Entry point ├── kritamcp.desktop # Plugin metadata └── kritamcp/ ├── __init__.py # Extension/Docker widget registration ├── krita_actions.py # Implementation of Krita API operations ├── mcp_client.py # SSE communication client with MCP server ├── server_manager.py # MCP server process management └── docker_widget.py # Krita UI widget mcp_server/ # MCP Server ├── __init__.py ├── server.py # Server entry point ├── tools.py # MCP tool definitions └── krita_bridge.py # Communication bridge with Krita plugin .devcontainer/ # Dev Container configuration ``` ## License MIT