# WDM AI Management: A Cross-Platform Unified Configuration Management Framework for AI Agents > An AI configuration management framework supporting GitHub Copilot CLI, Claude Code, OpenAI Codex CLI, and Gemini CLI, enabling 'write once, deploy everywhere' to solve the fragmentation issue of multi-AI assistant configurations. - 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm) - 发布时间: 2026-06-11T22:13:29.000Z - 最近活动: 2026-06-11T22:24:56.888Z - 热度: 141.8 - 关键词: AI配置管理, 多平台同步, Claude Code, GitHub Copilot, OpenAI Codex, Gemini CLI, 智能体管理, 开发工具标准化 - 页面链接: https://www.zingnex.cn/en/forum/thread/wdm-ai-management-ai - Canonical: https://www.zingnex.cn/forum/thread/wdm-ai-management-ai - Markdown 来源: floors_fallback --- ## Introduction to the WDM AI Management Framework ### Core Introduction to the WDM AI Management Framework WDM AI Management is an open-source cross-platform unified configuration management framework for AI agents, released by WD-Mitchell on GitHub on June 11, 2026. It supports tools like GitHub Copilot CLI, Claude Code, OpenAI Codex CLI, and Gemini CLI. Its core goal is to solve the fragmentation problem of multi-AI assistant configurations, enabling 'Write once, deploy everywhere' so that developers can manage all AI assistant configurations through a single source file. Project link: https://github.com/WD-Mitchell/WDM-AI-Management ## Background of the Dilemma in Multi-AI Assistant Configurations ### Fragmentation Issue of Multi-AI Assistant Configurations With the development of the AI programming assistant ecosystem, developers often use multiple tools simultaneously, but the configuration formats and storage locations of each tool vary significantly: - Claude Code: `.claude/agents/*.md` (YAML frontmatter includes model and effort fields) - Copilot CLI: `.github/agents/*.agent.md` (includes mcp-servers and user-invocable fields) - OpenAI Codex CLI: `~/.codex/agents/*.toml` (TOML format, includes model_reasoning_effort and sandbox_mode fields) - Gemini CLI: `.gemini/agents/*.md` (includes thinkingConfig.thinkingBudget field) Maintaining multiple sets of configurations is tedious and error-prone; modifications need to be synchronized across platforms, which easily leads to inconsistent experiences. ## Core Architecture and Configuration Methods ### Single Source File-Driven Multi-Platform Generation #### Architecture Workflow Single source file (YAML frontmatter + markdown body) → Generate platform-specific formats via the `wdm ai sync` command and complete deployment (symbolic links + configuration merging). #### Supported Content Types The framework supports 6 types of configurations: Agents, Skills, Rules, Workflows, Hooks, and MCP Servers. #### Field Override Mechanism - Basic fields: Shared across all platforms (name, description, model, etc.) - Platform prefix: `_` (e.g., copilot_model: claude-sonnet-4.6) - Multi-platform prefix: `

_

_` (e.g., codex_copilot_description) - Global fields: `global_` (unified value across all platforms) #### Model Layers Three layers are defined: default, default-small, and default-large, which map to platform-specific models and automatically inject reasoning parameters (e.g., Claude's effort, Codex's model_reasoning_effort). #### Deployment Modes - Global deployment (`-g`): Writes to the user's home directory and takes effect for all projects - Project-level deployment: Writes to the current project's subdirectory, supports version control, and facilitates team collaboration ## Technical Highlights and Usage Workflow ### Key Technologies and Operation Steps #### Technical Highlights 1. **Schema Whitelist**: Only retains fields defined in the official documentation of each tool to avoid unknown field leaks (implemented via AGENT_SCHEMAS in `ai_management/build.py`, etc.) 2. **Configuration Merging**: Uses a merging strategy for shared setting files (e.g., .codex/config.toml) to preserve users' existing configurations #### Usage Workflow 1. Installation: Install via npm, Bun, or Homebrew 2. Launch Web UI: Run `wdm ai` to start the local visual management interface 3. Write source file: Create a Markdown source file (YAML frontmatter defines metadata) 4. Deployment: `wdm ai --install` to install default groups; `wdm ai sync` for project-level deployment; `wdm ai sync -g` for global deployment 5. Preview: Use the `--dry-run` flag to preview changes without actually writing files ## Extensibility and Application Scenarios ### Framework Extensibility and Practical Value #### Extensibility Supports adding new AI platforms: Create a JSON definition file in the `harnesses/core/` directory, declaring the platform name, Schema, output extension, and synchronization target template to integrate it. #### Practical Application Scenarios - **Team Standardization**: Unify team AI assistant configurations to ensure consistent code reviews and security rules - **Personal Multi-Device Sync**: Keep AI configurations consistent across multiple machines - **Cross-Platform Migration**: No need to rewrite configurations when switching AI tools - **Version Control**: Include AI configurations in Git management to track change history ## Conclusion and Recommendations ### A Standardization Attempt for Configuration Management WDM AI Management is a standardization attempt for configuration management of AI development tools, solving the fragmentation issue of multi-AI assistant configurations through the 'single source file, multi-target generation' architecture. Recommendations for the following groups to try: - Developers who use multiple AI programming assistants simultaneously - Tech Leads who want to standardize AI configurations within their teams This tool provides an elegant solution for AI configuration management and is worthy of attention and use by the open-source community.