ClawX is a desktop application (macOS, Windows, and Linux) that puts a graphical interface over OpenClaw. It embeds the OpenClaw runtime inside the app rather than requiring a separate installation, and tracks upstream OpenClaw closely so new core capabilities carry over. A setup wizard and visual settings replace the command-line setup, configuration files, and process management that OpenClaw otherwise requires.
The app uses a dual-process design: an Electron main process owns the window, supervises the OpenClaw Gateway, handles system integration and updates, and bridges chat over the Agent Client Protocol (ACP). The React renderer talks to one client abstraction and never reaches local endpoints directly. Configuration changes such as providers, agents, skills, and models are applied without restarting the Gateway, and credentials are hot-reloaded.
Typical uses include a general desktop assistant, scheduled monitoring of news feeds or prices with results pushed to a messaging channel, and chaining skills into automation pipelines. ClawX disables OpenClaw's general-purpose web search tool and several internal agent tools at both the agent and Gateway policy layers, while managed browser automation and web fetching remain available.
Features
- Chat interface: multi-session history, streaming Markdown with syntax highlighting, KaTeX math, @agent routing, inline skill cards, subagent status drill-down, and previews for Markdown, DOCX, PPTX, and HTML
- Agent management: create and manage specialized agents from the desktop, each with its own managed workspace and chat history
- Multi-channel messaging: independent channels with multiple accounts and per-account agent binding, including bundled WeChat and DingTalk connectors
- Scheduled automation: recurring or one-time cron schedules that can insert skills into prompts and deliver results to external channels
- Skill system: local skill management independent of the Gateway, discovery from multiple OpenClaw sources, and bundled skills for PDF, XLSX, DOCX, and PPTX
- Provider integration: OpenAI, Anthropic, Z.AI / GLM, and custom providers, with credentials kept in the native system keychain
- Local computer use: on macOS and Windows x64, agents drive windows, menus, and accessibility controls through a bundled native Cua driver
- Voice dictation: a developer-mode microphone button that transcribes speech through a user-configured speech-to-text service
- Image generation: developer-mode image endpoints whose generated images render inline in chat
- Proxy settings: built-in proxy configuration for the app, the Gateway, and channels such as Telegram
- Issue report export: a ZIP bundle with sanitized OpenClaw configuration, diagnostic logs, and optionally selected transcripts
- Desktop conveniences: light, dark, or system themes, launch at startup, and update prompts before download
