Get OpenClaw-like personal AI agent in 5 mins. with Sandbox security,small and fast
English | 简体中文 | 繁體中文 | 日本語 | 한국어 | العربية | বাংলা | Deutsch | Español | Français | हिन्दी | Italiano | Português | Slovenščina | Türkçe | Tiếng Việt
Get OpenClaw-like knowledge base personal AI agent in 5 mins. Sandbox-secured, with an ultra-small 30MB installer for macOS & Windows (install in 1 min). Connects to WhatsApp, Telegram, Slack, Discord, Gmail, DingTalk, WeChat Work, QQ, Feishu & other messaging apps. Built-in Skill Market, Knowledge Base, Memory, MCP, Scheduled Tasks. Developed in Go: fast & low resource usage.
5分钟拥有类 OpenClaw 本地知识库个人AI智能体,沙箱安全防护,支持macOS/Windows 30M 极小安装包,1 分钟安装。连接WhatsApp、Telegram、Slack、Discord、Gmail、钉钉、企业微信、QQ、飞书等主流通讯应用，内置技能市场、IMA开源版本地知识库平替、记忆、MCP、计划任务等核心功能。Go语言开发，运行快、占资源少。
Ask your AI assistant any question, and it will intelligently search your knowledge base to generate a relevant answer.
Send a one-sentence command to the smart assistant to automatically create and generate a PowerPoint presentation.
Use a command to have the assistant help you find installed features on your computer or install new extension plugins.
Add external MCP servers to securely and efficiently connect to diverse data sources and tools, enabling your assistant to go beyond daily tasks into professional workflows.
Choose between sandbox-isolated execution (OS-level isolation, restricted command scope) and native execution (more flexible). Switch freely to balance safety and convenience.
Enable contextual conversations and personalized assistance. The assistant can continuously learn and evolve so it feels like a growing partner.
Authorize one-click access to ChatWiki to sync robots and knowledge bases, share configurations, and control member permissions.
Upload documents (TXT, PDF, Word, Excel, CSV, HTML, Markdown). The system automatically parses, splits, and converts them into vector embeddings for precise retrieval.
Integrate IM providers (Feishu, WeCom, QQ, DingTalk, LINE, Discord, WhatsApp, X/Twitter, Telegram, etc.) via SDKs to quickly enable channel creation, user management, and messaging.
Let your assistant automatically execute actions at preset times or intervals: reminders, recurring work, and system-level maintenance.
Select any text on your screen. It is automatically copied into a floating quick-ask box. One click to ask, instant answers.
Snap the assistant alongside other windows, quickly switch between differently configured assistants, and one-click send generated replies into your conversations.
Consult multiple "AI experts" simultaneously and view their responses side-by-side for easy comparison.
Click the floating ball on your desktop to instantly wake up or open the main ChatClaw window.
ChatClaw can run as a server (no desktop GUI required), accessible via a browser.
Download the binary for your platform from GitHub Releases:
| Platform | File |
|---|---|
| Linux x86_64 | ChatClaw-server-linux-amd64 |
| Linux ARM64 | ChatClaw-server-linux-arm64 |
chmod +x ChatClaw-server-linux-amd64
./ChatClaw-server-linux-amd64Open http://localhost:8080 in your browser.
The server listens on 0.0.0.0:8080 by default. You can customize host and port via environment variables:
WAILS_SERVER_HOST=127.0.0.1 WAILS_SERVER_PORT=3000 ./ChatClaw-server-linux-amd64docker run -d \
--name chatclaw-server \
-p 8080:8080 \
-v chatclaw-data:/root/.config/chatclaw \
registry.cn-hangzhou.aliyuncs.com/chatwiki/chatclaw:latestOpen http://localhost:8080 in your browser.
Create a docker-compose.yml file:
services:
chatclaw:
image: registry.cn-hangzhou.aliyuncs.com/chatwiki/chatclaw:latest
container_name: chatclaw-server
volumes:
- chatclaw-data:/root/.config/chatclaw
ports:
- "8080:8080"
restart: unless-stopped
volumes:
chatclaw-data:Then run:
docker compose up -dOpen http://localhost:8080 in your browser. To stop: docker compose down. Data is persisted in the chatclaw-data volume.
| Layer | Technology |
|---|---|
| Desktop Framework | Wails v3 (Go + WebView) |
| Backend Language | Go 1.26 |
| Frontend Framework | Vue 3 + TypeScript |
| UI Components | shadcn-vue (New York style) + Reka UI |
| Styling | Tailwind CSS v4 |
| State Management | Pinia |
| Build Tool | Vite |
| AI Framework | Eino (ByteDance CloudWeGo) |
| AI Model Providers | OpenAI / Claude / Gemini / Ollama / DeepSeek / Doubao / Qwen / Zhipu / Grok |
| Database | SQLite + sqlite-vec (vector search) |
| Internationalization | go-i18n + vue-i18n |
| Task Runner | Task |
| Icons | Lucide |
ChatClaw_D2/
├── main.go # Application entry point
├── go.mod / go.sum # Go module dependencies
├── Taskfile.yml # Task runner configuration
├── build/ # Build configurations & platform assets
│ ├── config.yml # Wails build config
│ ├── darwin/ # macOS build settings & entitlements
│ ├── windows/ # Windows installer (NSIS/MSIX) & manifests
│ ├── linux/ # Linux packaging (AppImage, nfpm)
│ ├── ios/ # iOS build settings
│ └── android/ # Android build settings
├── frontend/ # Vue 3 frontend application
│ ├── package.json # Node.js dependencies
│ ├── vite.config.ts # Vite bundler config
│ ├── components.json # shadcn-vue config
│ ├── index.html # Main window entry
│ ├── floatingball.html # Floating ball window entry
│ ├── selection.html # Text selection popup entry
│ ├── winsnap.html # Snap window entry
│ └── src/
│ ├── assets/ # Icons (SVG), images & global CSS
│ ├── components/ # Shared components
│ │ ├── layout/ # App layout, sidebar, title bar
│ │ └── ui/ # shadcn-vue primitives (button, dialog, toast…)
│ ├── composables/ # Vue composables (reusable logic)
│ ├── i18n/ # Frontend i18n setup
│ ├── locales/ # Translation files (zh-CN, en-US…)
│ ├── lib/ # Utility functions
│ ├── pages/ # Page-level views
│ │ ├── assistant/ # AI chat assistant page & components
│ │ ├── knowledge/ # Knowledge base management page
│ │ ├── multiask/ # Multi-model comparison page
│ │ └── settings/ # Settings page (providers, models, tools…)
│ ├── stores/ # Pinia state stores
│ ├── floatingball/ # Floating ball mini-app
│ ├── selection/ # Text selection mini-app
│ └── winsnap/ # Snap window mini-app
├── internal/ # Private Go packages
│ ├── bootstrap/ # Application initialization & wiring
│ ├── define/ # Constants, built-in providers, env flags
│ ├── device/ # Device identification
│ ├── eino/ # AI/LLM integration layer
│ │ ├── agent/ # Agent orchestration
│ │ ├── chatmodel/ # Chat model factory (multi-provider)
│ │ ├── embedding/ # Embedding model factory
│ │ ├── filesystem/ # File-system tools for AI agents
│ │ ├── parser/ # Document parsers (PDF, DOCX, XLSX, CSV)
│ │ ├── processor/ # Document processing pipeline
│ │ ├── raptor/ # RAPTOR recursive summarization
│ │ ├── splitter/ # Text splitter factory
│ │ └── tools/ # AI tool integrations (browser, search, calculator…)
│ ├── errs/ # i18n-aware error handling
│ ├── fts/ # Full-text search tokenizer
│ ├── logger/ # Structured logging
│ ├── services/ # Business logic services
│ │ ├── agents/ # Agent CRUD
│ │ ├── app/ # Application lifecycle
│ │ ├── browser/ # Browser automation (via chromedp)
│ │ ├── chat/ # Chat & streaming
│ │ ├── conversations/ # Conversation management
│ │ ├── document/ # Document upload & vectorization
│ │ ├── floatingball/ # Floating ball window (cross-platform)
│ │ ├── i18n/ # Backend internationalization
│ │ ├── library/ # Knowledge library CRUD
│ │ ├── multiask/ # Multi-model Q&A
│ │ ├── providers/ # AI provider configuration
│ │ ├── retrieval/ # RAG retrieval service
│ │ ├── settings/ # User settings with cache
│ │ ├── textselection/ # Screen text selection (cross-platform)
│ │ ├── thumbnail/ # Window thumbnail capture
│ │ ├── tray/ # System tray
│ │ ├── updater/ # Auto-update (GitHub/Gitee)
│ │ ├── windows/ # Window management & snap service
│ │ └── winsnapchat/ # Snap chat session service
│ ├── sqlite/ # Database layer (Bun ORM + migrations)
│ └── taskmanager/ # Background task scheduler
├── pkg/ # Public/reusable Go packages
│ ├── webviewpanel/ # Cross-platform webview panel manager
│ ├── winsnap/ # Window snapping engine (macOS/Windows/Linux)
│ └── winutil/ # Window activation utilities
├── docs/ # Development documentation
│ └── readmes/ # Multi-language README files
│ ├── README.md # English README
│ ├── README_zh-CN.md # Simplified Chinese README
│ └── images/ # README screenshots
└── images/ # README screenshots (legacy)
To view the complete update log, please click👉️👉️UpdateLog.md
- README Refresh (Multi-language): Updated README files across multiple languages with new previews, clearer capability descriptions, and consistent image paths/structure.
- Channel Integrations & Messaging: Improved messaging flows and guardrails for multiple channels (e.g., QQ config dedup & image sending, WeCom/Feishu streaming output, DingTalk checks and related config updates).
- Docs/Tooling: Added
readme-from-docxskill documentation for syncing README content from Word documents, including image extraction and localization guidelines.
- Chat File Upload (End-to-End): Delivered file attachment support across chat UI and backend services, including type/size validation, message state integration, and consistent handling alongside image attachments.
- Chat Input UX Upgrade: Enhanced
ChatInputAreawith a new conversation entry, dropdown integrations, a compact mode selector with icon toggles, and a bottom toolbar for better ergonomics. - In-App Selection Context Menu: Added an in-app text selection popup with actions and an option to disable selection search, with
TextSelectionServiceand settings synchronization. - MCP Defaults & Reliability: Enabled MCP by default via SQLite migration and hardened MCP settings flows (tool add/remove state handling, validation, cleanup) to reduce edge-case failures.
- Model Config Validation: Strengthened Azure chat/embedding configuration validation (endpoint/version requirements) and refined thinking feature initialization logic for safer defaults.
- Project Hygiene: Added
think_docs/convention +.gitignoreentry, removed legacy Klingon locale artifacts, and applied small refactors for readability/maintainability (e.g., button handlers).
- New Style / Knowledge UI Refactor: Major refactor of
KnowledgePagefolder/library behaviors with debounced expansion, improved folder-tree synchronization, and upgraded team knowledge UI viaTeamFolderCard/TeamFileCardcomponents. - Assistant MCP Feature Expansion: Added an Assistant MCP detail view (edit + tool management) and improved service networking behavior (host binding checks,
127.0.0.1usage, CORS/OPTIONS handling). - i18n Loading Improvements: Refactored i18n initialization to support async locale message loading and made message typing more flexible for dynamic locale structures.
- Performance & Build Optimizations: Improved Vite bundling via manual chunking and async component loading in
App.vue; added reusablecopyToClipboardutility; tuned server-mode performance and tokenizer dictionary loading for Chinese segmentation. - Build/Dev Workflow Updates: Updated
development.mdand Dockerfiles for clearer, more reliable build steps (bindings generation, frontend deps), plus backend app init/systray refinements for stability.
