A local-first AI workspace built on Rust and Tauri, acting as an AI coworker for secure, supervised automation on any folder. Supports multiple LLM backends, MCP protocol extension, and multimodal file processing.
Project Positioning#
Tandem is a local-first, cross-platform AI workspace positioned as an "AI coworker" that performs secure, supervised automation on any folder.
Core Features#
Privacy & Security#
- Zero Telemetry: No usage data leaves the local machine except user-initiated external LLM API calls
- API Key Encryption: AES-256-GCM encrypted storage in local secure vault
- Zero Trust File Access: Default-deny for sensitive files (.env, .pem, .key, SSH keys), all operations require explicit approval
- Sandboxed Execution: Based on Tauri security model with strict CSP and permission controls
AI Agent Modes#
- Chat Mode: Interactive dialogue with context-aware file operations
- Plan Mode: Generate detailed implementation plans (.md format) before execution, supports batch review and rollback
- Ralph Loop: Autonomous iteration loop until task is verifiably complete
- Ask Mode: Read-only exploration and analysis mode
- Debug Mode: Runtime evidence-based systematic debugging
- Multi-Agent Orchestration: Coordinates specialized sub-agents (Planner/Builder/Validator) for complex problems
Provider Flexibility#
Supports OpenRouter, Anthropic, OpenAI, Ollama, and any OpenAI-compatible API. Provider-agnostic design avoids vendor lock-in.
Integration & Extension#
- MCP Connectors: Supports Model Context Protocol for connecting to any MCP server
- External Channels: Slack, Discord, Telegram integration
- Routine Scheduler: Supports scheduled automation tasks
- Skills System: Supports custom AI capabilities and instruction imports
File Processing & Output#
- Format Parsing: PDF, DOCX, PPTX, XLSX, RTF text extraction
- Output Artifacts: HTML/Canvas interactive dashboards, PPTX presentations, Markdown reports
Architecture#
- Frontend: React 18, TypeScript, Tailwind CSS, Framer Motion
- Backend: Rust, Tauri 2.0
- Agent Runtime: Tandem Engine (Rust, HTTP + SSE)
- Vector Database: sqlite-vec (long-term memory support)
Installation#
Prerequisites#
- Node.js 20+
- Rust 1.75+ (via rustup)
- pnpm (recommended) or npm
Platform Dependencies#
- Windows: Build Tools for Visual Studio
- macOS: Xcode Command Line Tools
- Linux: libwebkit2gtk-4.1-dev, libappindicator3-dev, librsvg2-dev, build-essential, pkg-config
Build from Source#
git clone https://github.com/frumu-ai/tandem.git
cd tandem
pnpm install
cargo build -p tandem-ai
pnpm tauri dev
NPM Package Installation#
# TUI terminal interface
pnpm add -g @frumu/tandem-tui
# Headless engine core
pnpm add -g @frumu/tandem
Typical Use Cases#
| Scenario | Description |
|---|---|
| Codebase Intelligence | Full codebase understanding, open-source and provider-agnostic |
| Research & Document Processing | Extract text from PDF/DOCX/PPTX/XLSX, synthesize research papers |
| Multi-step Automation | Decompose complex tasks into reviewable steps, batch approval |
| Change Review | Visualize file changes before/after, one-click rollback |
| Cross-functional Team Support | Specialized skill packages for marketing, legal, sales, data science |
Project Info#
- Primary Languages: Rust (48.5%), TypeScript (19.3%), Python (18.1%), HTML (9.4%)
- Platform Support: Windows, macOS (Intel & Apple Silicon), Linux
- Repository Activity: 502 commits, 57 releases (as of v0.3.19)