# ArtClaw Bridge **Bridge DCC tools (UE, Maya, 3ds Max, Blender, Houdini, Substance Painter/Designer, ComfyUI) to AI Agents via MCP Protocol** > ๐ŸŽฏ **The Missing Bridge Between AI Agents and Creative Software** ArtClaw Bridge provides a unified AI bridging layer for Digital Content Creation (DCC) software. Through the [MCP (Model Context Protocol)](https://modelcontextprotocol.io/), AI Agents can directly understand and operate the editor environment โ€” enabling AI to truly integrate into game development, film & animation, and digital art workflows. **Keywords:** `maya-mcp` `blender-mcp` `unreal-engine-mcp` `3dsmax-mcp` `houdini-mcp` `substance-mcp` `comfyui-mcp` `maya-skill` `blender-skill` `ue-skill` `ai-dcc` `ai-3d` `mcp-bridge` `ai-game-dev` `ai-vfx` ![License](https://img.shields.io/badge/license-MIT-blue.svg) ![Status](https://img.shields.io/badge/status-beta-orange.svg) ![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg) [ไธญๆ–‡ๆ–‡ๆกฃ](README_zh.md) --- ## ๐ŸŽฏ Why ArtClaw Bridge? ### The Problem We're Solving **1. AI Agents Can't Directly Operate Creative Software** Most AI agents (OpenClaw, Claude Code, Cursor, etc.) are general-purpose platforms. They excel at coding and text tasks, but they don't know how to: - Select objects in Maya - Create materials in Unreal Engine - Generate textures in Substance Designer - Run workflows in ComfyUI **2. From Custom Tools to Tool-Creation Platform** Traditionally, every DCC required custom scripts and tools built one-by-one. ArtClaw Bridge shifts this paradigm: - **Before**: Each tool is hand-crafted code, requiring deep expertise in the DCC's API - **After**: Artists describe what they need in natural language, AI generates the tool automatically The bridge layer handles the DCC-specific complexity once. After that, tool creation becomes a conversation, not a coding task. **3. Tool Management for the AI Era** Traditional tool frameworks were designed for manual discovery and learning. ArtClaw's Tool Manager reimagines this: - **AI-Native Discovery**: No more hunting through menus โ€” just ask AI to find and run the right tool - **Zero Learning Curve**: AI explains and executes tools on-demand, eliminating training time - **Team Sharing**: Skills and tools are versioned, shared, and synced across the team - **Personal + Team Workflows**: Individuals create tools for their needs; teams curate shared libraries The result: tools are created, shared, discovered, and executed through AI โ€” a framework built for how people actually work with AI. ### Our Solution ArtClaw Bridge is a **unified bridging layer** that: - โœ… Connects any MCP-compatible AI Agent to any supported DCC software - โœ… Provides a standardized protocol (MCP) for AI-to-software communication - โœ… Includes an AI-powered Tool Manager for natural language tool creation - โœ… Works with existing Agent platforms โ€” no need to build yet another Agent system > ๐Ÿ’ก **Think of it as the "USB-C for AI Agents and Creative Software"** โ€” one bridge, infinite combinations. --- ## ๐ŸŽฌ Demo > Real operation demos showing AI Agent executing tasks directly in the editor **UE Connection to OpenClaw โ€” AI Chat Panel in Editor** ![UEClawBridge](docs/็คบไพ‹/UE่ฟžๆŽฅopenClawๆ•ˆๆžœ.png) **Blender Integration โ€” AI Operating in Blender Editor** ![Blender AI Integration](docs/็คบไพ‹/blender-bridgeๆผ”็คบ.gif) **Cross-DCC Pipeline โ€” AI-Driven Workflow** ![ArtClawBridge Workflow](docs/็คบไพ‹/ๆ‰“้€šไธๅŒ่ฝฏไปถ้—ด็š„ไธŠไธ‹ๆตไบคๆŽฅ.png) **Substance Designer & Painter โ€” AI Texture Generation Workflow** ![SD AI Integration](docs/็คบไพ‹/SDๆŽฅๅ…ฅ.png) ![SP AI Integration](docs/็คบไพ‹/SPๆŽฅๅ…ฅ.png) **Tool Manager โ€” Web Skillใ€Tool Manager** ![Web-based chat panel - Quick operation with tools - AI-assisted operation](docs/็คบไพ‹/tool-manager-chat้ขๆฟ-ๅทฅๅ…ทๅฟซๆท่ฟ่กŒ-AIๅๅŠฉ่ฟ่กŒ.png) ![skill management - Team sharing - Category filtering](docs/็คบไพ‹/tool-manager-skill็ฎก็†-ๅ›ข้˜Ÿๅ…ฑไบซ-ๅˆ†็ฑป็ญ›้€‰.png) ![Tool Management - Filtering Conditions + Trigger Rules Paradigm](docs/็คบไพ‹/tool-manager-tool็ฎก็†-็ญ›้€‰ๆกไปถ+่งฆๅ‘่ง„ๅˆ™-่Œƒๅผ.png) โญ *More demo videos coming soon!* --- ## โœจ Core Features ### ๐Ÿ”— Unified MCP Protocol All DCC software communicates with AI Agents through the standard MCP protocol. Each DCC exposes only one MCP tool (`run_ue_python` / `run_python`), and AI completes all operations by executing Python code โ€” minimal yet powerful. ### ๐Ÿ’ฌ In-Editor AI Chat Panel Chat directly with AI in UE / Maya / Max / Blender / Houdini / SP / SD / ComfyUI editors without switching windows. Features: - **Streaming Output** โ€” AI responses display in real-time with Markdown rendering - **Tool Call Visualization** โ€” Collapsible cards showing tool name, parameters, and execution results - **Attachment Support** โ€” Drag and drop images or files for AI to analyze automatically - **Context Length Display** โ€” Real-time token usage percentage - **Stop Button** โ€” Interrupt AI execution anytime (sends `chat.abort` to terminate Agent) ### ๐Ÿ› ๏ธ Skill Management System Layered Skill hot-reloading system, shared across DCCs: - **Four-Level Priority** โ€” Official > Marketplace > User > Temporary, higher levels override lower ones - **In-Editor Management Panel** โ€” Unified UI for both UE + DCC, supports filter/search/enable/disable/pin - **Full Lifecycle** โ€” Install, uninstall, update, publish (version increment + git commit), one-click full sync - **AI-Generated Skills** โ€” Describe requirements in natural language, AI auto-generates executable Skills (manifest + code + docs) - **Change Detection** โ€” Auto-detects unpublished changes at runtime, intelligently distinguishes "update" vs "publish" direction - **Pinned Skills Context Injection** โ€” Pinned Skill docs automatically injected into AI's first message context ### ๐ŸŒ Multi-Agent Platform Support Configuration-driven platform abstraction layer โ€” new platforms automatically appear in UI when registered in config: - **OpenClaw** โ€” Primary platform, integrated via mcp-bridge plugin (Full mode: in-editor chat + MCP) - **LobsterAI (Youdao)** โ€” OpenClaw repackaged, Gateway port 18790 (Full mode) - **Claude Code / Cursor / WorkBuddy** โ€” MCP-only mode via stdio bridge, use DCC tools from IDE/terminal - **Hot-Swap in Editor** โ€” One-click platform switch in Settings panel, auto disconnect/reconnect/refresh Agent list ### ๐Ÿ”„ Multi-Session & Agent Management - **Multi-Agent Switching** โ€” Select Agent in settings panel, toolbar shows current Agent info - **Session List Management** โ€” Create/switch/delete conversations, each Agent has independent session cache - **Session Persistence** โ€” Auto-recover last session after UE restart, DCC saves session state in real-time ### ๐Ÿง  Memory Management System v2 Three-tier progressive memory model โ€” AI remembers user preferences and operation history: - **Short-term** (4h / 200 entries) โ†’ **Medium-term** (7d / 500 entries) โ†’ **Long-term** (permanent / 1000 entries) - Semantic tag classification (facts/preferences/norms/operations/crashes/patterns) - Auto-promotion, deduplication, scheduled maintenance - Operation history tracking and querying ### ๐Ÿ“š Local Knowledge Base (RAG) Index API docs and project docs, semantic retrieval assists AI decision-making. ### ๐Ÿ›ก๏ธ Security & Stability - Transaction protection, risk assessment, main-thread scheduling - Shared module sync verification (`verify_sync.py` compares MD5, prevents multi-copy drift) - Long-task timeout protection + active event reset --- ## ๐ŸŽฏ Supported Engines, DCCs & Agent Platforms Currently verified with **OpenClaw + LobsterAI + Unreal Engine 5.7 + Maya 2023 + 3ds Max 2023 + Blender 5.1 + Substance Painter 11.0.1 + Substance Designer 12.1.0 + ComfyUI**. MCP-only platforms (Claude Code, Cursor, WorkBuddy) are configured and deployable. Other combinations are theoretically compatible but not tested โ€” community feedback welcome. ### Engines & DCC Software | Software | Verified Version | Status | Plugin | MCP Port | Notes | |----------|-----------------|--------|--------|----------|-------| | **Unreal Engine** | 5.7 | โœ… Verified | UEClawBridge | 8080 | C++ + Python, Slate UI chat panel | | **Maya** | 2023 | โœ… Verified | DCCClawBridge | 8081 | Python 3.9.7 + PySide2, Qt chat panel | | **3ds Max** | 2023 | โœ… Verified | DCCClawBridge | 8082 | Python 3.9.7 + PySide2, Qt chat panel, MaxScript startup | | **Blender** | 5.1 | โœ… Verified | DCCClawBridge | 8083 | PySide6 standalone Qt window, bpy.app.timers driven | | **Houdini** | โ€” | โš ๏ธ Not Verified | DCCClawBridge | 8084 | Code implemented, hdefereval main-thread scheduling, not tested | | **Substance Painter** | 11.0.1 | โœ… Verified | DCCClawBridge | 8085 | SP built-in Qt + QTimer polling | | **Substance Designer** | 12.1.0 | โœ… Verified | DCCClawBridge | 8086 | SD built-in Qt + QTimer polling, pre-injected sd.api vars | | **ComfyUI** | V0.19.0 | โœ… Verified | ComfyUIClawBridge | 8087 | Custom node, Python-only, WS MCP Server, no visible nodes | | **Other UE / Maya Versions** | โ€” | โš ๏ธ Not Verified | โ€” | โ€” | Theoretically compatible with UE 5.3+ / Maya 2022+, not tested | ### Agent Platforms | Platform | Mode | Status | Notes | |----------|------|--------|-------| | **OpenClaw** | Full (Gateway) | โœ… Verified | Primary dev platform, in-editor chat + MCP tools, all features | | **LobsterAI (Youdao)** | Full (Gateway) | โœ… Verified | OpenClaw repackaged, Gateway port 18790, in-editor chat + MCP tools | | **Claude Code** | MCP-only (stdio) | โœ… Configured | CLI tool, MCP tools in terminal, no in-editor chat | | **Cursor** | MCP-only (stdio) | โœ… Configured | VSCode fork, MCP tools in IDE, no in-editor chat | | **WorkBuddy (Tencent)** | MCP-only (stdio) | โœ… Configured | VSCode fork, MCP tools in IDE, no in-editor chat | > **Full mode** = in-editor AI chat panel + MCP tool calls (requires Gateway). > **MCP-only mode** = MCP tool calls only, chat happens in the platform's own UI (via `artclaw_stdio_bridge.py`). --- ## ๐Ÿ› ๏ธ Official Skills (47 Total) ### Universal Skills (5) - **artclaw-knowledge** โ€” Project knowledge base queries - **artclaw-memory** โ€” Memory management operations - **artclaw-skill-manage** โ€” Skill management operations - **artclaw-tool-creator** โ€” AI-guided tool creation - **artclaw-tool-executor** โ€” Tool execution dispatcher ### Unreal Engine Skills (6) - **ue57-artclaw-context** โ€” Editor context queries - **ue57-artclaw-highlight** โ€” Viewport highlighting - **ue57-camera-transform** โ€” Camera operations - **ue57-operation-rules** โ€” UE operation guidelines - **ue57_get_material_nodes** โ€” Material node queries - **ue57_material_node_edit** โ€” Material node editing ### Maya Skills (1) - **maya-operation-rules** โ€” Maya operation guidelines ### Blender Skills (4) - **blender-context** โ€” Editor context queries - **blender-material-ops** โ€” Material & node operations - **blender-operation-rules** โ€” Blender operation guidelines - **blender-viewport-capture** โ€” Viewport screenshot ### 3ds Max Skills (1) - **max-operation-rules** โ€” Max operation guidelines ### Houdini Skills (4) - **houdini-context** โ€” Editor context queries - **houdini-node-ops** โ€” Node operations - **houdini-operation-rules** โ€” Houdini operation guidelines - **houdini-simulation** โ€” Simulation operations ### Substance Painter Skills (4) - **sp-context** โ€” Editor context queries - **sp-layer-ops** โ€” Layer operations - **sp-operation-rules** โ€” SP operation guidelines - **sp-bake-export** โ€” Baking and export ### Substance Designer Skills (9) - **sd-context** โ€” Editor context queries - **sd-learned-recipes** โ€” Material recipe library - **sd-node-ops** โ€” Node operations - **sd-node-catalog** โ€” Node catalog reference - **sd-node-capture** โ€” Node screenshot & analysis - **sd-operation-rules** โ€” SD operation guidelines - **sd-generators** โ€” Generator node reference - **sd-fxmap** โ€” FX-Map creation guide - **sd-pixel-processor** โ€” Pixel Processor guide ### ComfyUI Skills (13) - **comfyui-operation-rules** โ€” ComfyUI operation guidelines - **comfyui-context** โ€” System info, models, queue status - **comfyui-workflow-builder** โ€” Workflow JSON construction guide - **comfyui-txt2img** โ€” Text-to-image workflow - **comfyui-img2img** โ€” Image-to-image workflow - **comfyui-inpainting** โ€” Inpainting workflow - **comfyui-controlnet** โ€” ControlNet workflow - **comfyui-hires-fix** โ€” High-resolution fix workflow - **comfyui-model-manager** โ€” Model management - **comfyui-node-installer** โ€” Missing node detection & install - **comfyui-workflow-repair** โ€” Workflow diagnosis & repair - **comfyui-workflow-manager** โ€” Workflow template library - **comfyui-workflow-validator** โ€” Workflow validation --- ## ๐Ÿ—๏ธ Architecture ``` โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”‚ AI Agent (LLM) โ”‚ โ”‚ OpenClaw / LobsterAI / Claude Code / Cursor โ”‚ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ”‚ WebSocket (Upstream: Chat RPC / Downstream: MCP Tool Calls) โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ–ผโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”‚ Agent Gateway โ”‚ โ”‚ + MCP Bridge + ArtClawToolManager โ”‚ โ”‚ (Unified Agent, Session, MCP Server, Tool, Workflow mgmt) โ”‚ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ”‚ WebSocket JSON-RPC (MCP) โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ผโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ–ผ โ–ผ โ–ผ โ–ผ โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”‚ UE โ”‚ โ”‚ Maya โ”‚ โ”‚ Blender โ”‚ โ”‚ ComfyUI โ”‚ โ”‚ :8080 โ”‚ โ”‚ :8081 โ”‚ โ”‚ :8083 โ”‚ โ”‚ :8087 โ”‚ โ””โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”ฌโ”€โ”€โ”€โ”€โ”˜ โ–ผ โ–ผ โ–ผ โ–ผ UE API Maya API bpy API ComfyUI API โ”‚ โ”‚ โ”‚ โ”‚ โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”Œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ” โ”‚ 3dsMax โ”‚ โ”‚ Houdini โ”‚ โ”‚ SP โ”‚ โ”‚ SD โ”‚ โ”‚ :8082 โ”‚ โ”‚ :8084 โ”‚ โ”‚ :8085 โ”‚ โ”‚ :8086 โ”‚ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ โ””โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”˜ ``` **Dual-Link Communication**: - **Upstream (Chat)**: Editor Panel โ†’ Gateway WebSocket RPC โ†’ AI Agent - **Downstream (Tool Calls)**: AI Agent โ†’ Gateway โ†’ MCP Bridge โ†’ DCC MCP Server โ†’ DCC API Each DCC software runs an independent MCP Server, exposing editor capabilities to AI Agents through a unified protocol. Skill system, knowledge base, memory storage, and other core modules are shared across DCCs. --- ## ๐Ÿ–ฅ๏ธ ArtClawToolManager โ€” Web Management Dashboard A standalone web-based management interface for unified control of Skills, Tools, and Workflows across all DCC platforms. ### Features - **Web-Based Chat Panel** โ€” Control AI Agents directly from the browser, no DCC installation required - **Skill Management** โ€” Browse, install, update, and manage all Skills across platforms - **Tool Registry** โ€” Unified registry for content filtering and automation tools - **Workflow Management** โ€” Create, manage, and execute AI-driven workflow templates - **ComfyUI Integration** โ€” Direct AI control of ComfyUI through the web interface - **Cross-Platform** โ€” Single dashboard manages all connected DCC software ### Usage The web page panel is connected to the same Agent Gateway as the DCC plugin, providing an alternative interface for AI conversations and tool/workflow management. It needs to be used in conjunction with the DCC software. --- ## ๐Ÿ”ง Tool & Workflow Architecture A unified paradigm designed for **content filtering, automation triggering, and AI-driven workflows**. Built as a three-layer architecture: ### Tool Layer - **Registration** โ€” Tools register with the ToolManager via standard manifest - **Execution** โ€” Unified `execute_tool(tool_id, inputs)` API with standardized I/O - **Filtering** โ€” Built-in content filtering pipeline (validate โ†’ filter โ†’ transform โ†’ output) - **Scheduling** โ€” Cron-based and event-triggered automation ### Workflow Layer - **Template System** โ€” JSON-based workflow templates with variable substitution - **Chain Execution** โ€” Multi-step workflows with dependency resolution - **State Management** โ€” Workflow state persistence across sessions - **AI Integration** โ€” LLMs can trigger workflows via `execute_workflow()` API ### Trigger Layer - **Event-Driven** โ€” File system events, API calls, message queues - **Conditional Logic** โ€” Filter conditions, threshold checks, content validation - **Fan-Out** โ€” Single trigger โ†’ multiple tool/workflow execution - **Audit Trail** โ€” Full execution logging for debugging and compliance This architecture enables powerful automation scenarios: - AI detects scene changes โ†’ triggers content validation workflow - New asset imported โ†’ runs automated QA tools - ComfyUI generation complete โ†’ triggers downstream DCC pipeline --- ## ๐Ÿš€ Installation > ๐Ÿ’ก **The Easiest Way: Let AI Install It For You** Instead of manual installation, simply tell your AI Agent: > *"Install ArtClaw Bridge for me. I need it for [UE/Maya/Blender/ComfyUI/etc.] at [path if needed]."* Your Agent will handle everything โ€” cloning, dependencies, configuration, and MCP setup. > ๐Ÿค– **AI Agents: Read [INSTALL_GUIDE.md](INSTALL_GUIDE.md) for step-by-step installation instructions.** It covers prerequisites, CLI commands, platform configuration, and post-install verification. **Example prompts:** - *"Install ArtClaw Bridge for Unreal Engine 5.7, my project is at D:\MyProject\UE_Game"* - *"Set up ArtClaw Bridge for Maya 2023 and Blender 5.1"* - *"Install ArtClaw Bridge for ComfyUI at C:\ComfyUI"* - *"Install ArtClaw Bridge with all DCC support"* ### Prerequisites - **Python** 3.9+ - **Agent Platform** (choose one): - [OpenClaw](https://github.com/openclaw/openclaw) (`npm install -g openclaw`) โ€” Full mode - [LobsterAI](https://lobsterai.com/) (Youdao) โ€” Full mode - [Claude Code](https://claude.ai/code) / [Cursor](https://cursor.com/) / [WorkBuddy](https://codebuddy.ai/) โ€” MCP-only mode - Target DCC software (choose as needed): - UE 5.7 (recommended, theoretically compatible with 5.3+) - Maya 2023 (recommended, theoretically compatible with 2022+) - 3ds Max 2023 (verified, theoretically compatible with 2020+) - Blender 5.1 (verified, auto-installs PySide6) - Houdini (not tested) - Substance Painter 11.0.1 (verified) - Substance Designer 12.1.0 (verified) - ComfyUI (verified, install via ComfyUI-Manager or copy to custom_nodes) ### Manual Installation (If You Prefer) ```bash # 1. Clone repo git clone https://github.com/IvanYangYangXi/artclaw_bridge.git cd artclaw_bridge # 2. Run interactive installer install.bat # Windows # or python install.py --help # View all options ``` The installer will automatically: 1. Link plugin files to target DCC standard directories via junction/symlink (fallback to copy if linking fails; use `--copy` to force copy mode) 2. Deploy `core/` shared modules (self-contained, no source directory needed) 3. Install official Skills to platform directory (`~/.openclaw/workspace/skills/` or LobsterAI equivalent) 4. **Safely handle startup files** (append mode, doesn't overwrite existing user content) 5. Configure Agent platform mcp-bridge integration 6. Write `~/.artclaw/config.json` project config 7. Idempotent (safe to run multiple times) For detailed manual installation instructions, see [INSTALL_GUIDE.md](INSTALL_GUIDE.md). ### Post-Install Verification | DCC | Verification Steps | |-----|-------------------| | **UE** | Open project โ†’ Enable "UE Claw Bridge" plugin โ†’ Restart โ†’ Window โ†’ UE Claw Bridge โ†’ Connect | | **Maya** | Launch Maya โ†’ **ArtClaw** appears in menu bar โ†’ Open Chat Panel โ†’ Connect | | **3ds Max** | Launch Max โ†’ ArtClaw auto-loads โ†’ Menu bar ArtClaw โ†’ Chat Panel โ†’ Connect | | **Blender** | Launch Blender โ†’ Edit โ†’ Preferences โ†’ Add-ons โ†’ Enable ArtClaw Bridge โ†’ Sidebar ArtClaw โ†’ Start Bridge | | **Houdini** | Launch Houdini โ†’ Shelf โ†’ ArtClaw โ†’ Start Bridge | | **SP** | Launch Substance Painter โ†’ Python โ†’ artclaw โ†’ start_plugin โ†’ Chat Panel | | **SD** | Launch Substance Designer โ†’ Python โ†’ artclaw โ†’ start_plugin โ†’ Chat Panel | | **ComfyUI** | Launch ComfyUI โ†’ Check console for "ArtClaw: ComfyUI Bridge ๅฏๅŠจไธญ..." โ†’ Web dashboard detects connection | ### Uninstall ```bash python install.py --uninstall --maya # Uninstall Maya plugin python install.py --uninstall --ue --ue-project "C:\project" # Uninstall UE plugin python install.py --uninstall --blender --blender-version 5.1 # Uninstall Blender plugin python install.py --uninstall --sp # Uninstall Substance Painter plugin python install.py --uninstall --sd # Uninstall Substance Designer plugin python install.py --uninstall --comfyui --comfyui-path "C:\ComfyUI" # Uninstall ComfyUI plugin ``` The uninstall script removes plugin directories and **only removes ArtClaw code blocks** from startup files (doesn't affect existing user content). --- ## ๐Ÿ› ๏ธ Skill System ### Directory Structure **Workflow**: Edit installed directory โ†’ `Publish` (installedโ†’source + version increment + git commit) โ†’ `Update` (sourceโ†’installed) ### Creating Skills Describe in natural language directly in the editor: > "Create a skill for me to batch rename selected Actors in the scene, adding a specified prefix" AI will auto-generate `SKILL.md` + `manifest.json` + `__init__.py`, ready to use after confirmation. --- ## ๐Ÿค Contributing Issues and Pull Requests welcome! Especially looking for contributions in: - ๐Ÿ”Œ **New DCC Bridge Implementations** โ€” Support for more DCC software - ๐Ÿ› ๏ธ **New Skills** โ€” Useful Skills for various DCCs (currently have UE / Maya / Max / Blender / Houdini / SP / SD / ComfyUI official Skills) - ๐Ÿงช **Testing Feedback** โ€” Test on unverified DCC versions and report - ๐Ÿ“– **Documentation** โ€” Usage tutorials, best practices ### Contribution Workflow 1. Fork this repository 2. Create feature branch: `git checkout -b feat/my-feature` 3. Commit changes: `git commit -m "feat: add my feature"` 4. Push and create PR See [Contributing Guide](docs/skills/CONTRIBUTING.md) for details. --- ## ๐Ÿ“– Documentation - **[System Architecture](docs/specs/็ณป็ปŸๆžถๆž„่ฎพ่ฎก.md)** โ€” Overall architecture and design principles - **[Skill Development Guide](docs/skills/SKILL_DEVELOPMENT_GUIDE.md)** โ€” Writing custom Skills - **[Skill Specification](docs/skills/MANIFEST_SPEC.md)** โ€” manifest.json format specification - **[Code Standards](docs/specs/ไปฃ็ ่ง„่Œƒ.md)** โ€” Project coding conventions - **[Multi-Platform Compatibility](docs/UEClawBridge/features/ๅคšๅนณๅฐๅ…ผๅฎน่ฎพ่ฎกๆ–นๆกˆ.md)** โ€” Platform abstraction layer design - **[DCCClawBridge](subprojects/DCCClawBridge/README.md)** โ€” Maya / Max / Blender / Houdini / SP / SD plugin details - **[ComfyUIClawBridge](subprojects/ComfyUIClawBridge/__init__.py)** โ€” ComfyUI plugin details - **[Contributing Guide](docs/skills/CONTRIBUTING.md)** โ€” How to contribute --- ## ๐Ÿงพ Design Philosophy ### Why Not Build Yet Another Agent Platform? Agent platforms are a big undertaking. Many companies are already building excellent ones (OpenClaw, LobsterAI, Claude Code, Cursor, etc.). This project focuses on **the specific engineering problem of bridging** โ€” enabling these existing platforms to control creative software. We don't reinvent the wheel; we make the wheel fit more vehicles. ### Why MCP + Skills Instead of Direct LLM Integration? Direct LLM integration is a black box โ€” unpredictable and uncontrollable. Like early AI image generation: impressive, but not production-ready. ArtClaw Bridge adds **engineering discipline** to AI execution: - Structured Skills provide predictable behavior - MCP protocol standardizes communication - Tool Manager enables controllable automation This approach mirrors Claude Code's philosophy: use engineering to make LLM execution correct and controllable, not magical. ### Who Benefits? - **Artists** โ€” Can leverage AI for automation without learning to code - **Technical Artists** โ€” Can build and share reusable Skills across teams - **Developers** โ€” Can integrate AI into existing pipelines without building custom Agent infrastructure - **Studios** โ€” Can standardize AI tooling across multiple DCCs and projects --- ## ๐Ÿ“„ License This project is open-sourced under [MIT License](LICENSE). ## ๐Ÿ‘ค Author **Ivan (Yang Jili)** โ€” [@IvanYangYangXi](https://github.com/IvanYangYangXi) --- ## โ˜• Support This Project If ArtClaw Bridge helps your work, consider buying the author a coffee โ˜• [![GitHub Sponsors](https://img.shields.io/github/sponsors/IvanYangYangXi?style=social)](https://github.com/sponsors/IvanYangYangXi) Your support is the biggest motivation for continued development and maintenance! ๐Ÿš€