An open-source, local-first desktop AI Agent workstation that transforms files, conversations, and office context into deliverables, positioned as a BYOK alternative to Claude Code and OpenAI Codex.
Core Capabilities#
File & Data Understanding#
- Native support for DOCX, XLSX, PPTX, PDF, CSV formats, plus local folders and generated Artifacts
- All data parsing happens locally, no upload to third-party servers
Workspace & Output#
- Artifact Panel: Reusable Markdown briefs, tables, charts, checklists displayed alongside conversations
- Workspace Memory: Project-level context stored as Markdown documents, auto-injected per session
Tool Execution & Automation#
- 20+ file operation tools (read/write, rename, organize), all requiring user authorization
- Automated scheduling for periodic cleanup, report generation, and file workflows
Context & Communication#
- Continuous conversations from analysis to planning to follow-up, with built-in context auto-compression
- 10+ built-in IM channels: WhatsApp, Telegram, Discord, Slack, WeChat, Feishu, DingTalk, Enterprise WeChat, QQ, Email, Matrix, MoChat, WebSocket
Privacy & Connectivity#
- Fully offline local model support, local SQLite storage, zero telemetry
- Secure remote access via Cloudflare Tunnel with QR code scanning from mobile devices
Architecture#
Monorepo structure:
desktop-tauri/: Tauri v2 (Rust) desktop shell for system integration and cross-platform packagingfrontend/: Next.js 15 UI with chat interface, settings, Artifact panel, SSE streamingbackend/: FastAPI Agent engine for tool execution, LLM streaming routing, SQLite persistence
Key mechanisms:
- FastAPI provides core Agent capabilities and SSE streaming API; Next.js handles interaction; Tauri packages as native desktop app
- OpenRouter aggregates 100+ models, compatible with 13+ cloud provider APIs and local Ollama auto-detection
- MCP (Model Context Protocol) standard for external tool integration, 46 pre-configured connectors with one-click enable and OAuth
Model Support#
- Cloud/Subscription: OpenRouter (free & paid), OpenAI, Anthropic, Google Gemini, DeepSeek, Groq, Mistral, xAI Grok, Qwen, Kimi Moonshot, MiniMax, Zhipu, ChatGPT subscription
- Local: Ollama auto-detection, offline capable
Typical Use Cases#
- Office Automation: Batch rename, organize, clean files across folders with auditable logs
- Data Analysis: Local XLSX/CSV/document parsing for anomaly detection, trend tracking, summary export
- Team Operations: Merge notes from PDF/DOCX/CSV into structured briefs with auto-extracted action items
- Content & Writing: Draft internal notices, policy updates, professional communications
- Remote Workflows: Send tasks to desktop instance from any device via secure tunnel
- IM Integration: Connect desktop AI Agent to mainstream IM platforms for automated message handling
Installation#
- macOS (Apple Silicon / Intel): .dmg, .app
- Windows (x64): .exe installer
- Linux (x64): .deb, .rpm
Dev setup: git clone https://github.com/openyak/openyak.git && cd openyak && npm run dev:all
Pending Confirmation#
- Core team/author info not explicitly listed on GitHub or website
- Pricing page exists but specific pricing details not obtained
- IM channel integration depth needs further verification
- Full list of 46 MCP connectors not fully presented
- Specific automation scheduling mechanism (cron/built-in scheduler) not clarified