🪔 Artha

Your work, done. Locally.

A local-first AI agent for document workflows, MCP tools, and agentic automation.
No data leaves your machine. Ever.

---

What is Artha?

Artha (Sanskrit: अर्थ — purpose, livelihood, work) is a local-first AI agent desktop app that lets you describe any task in plain English and have it executed, produced, and delivered entirely on your machine.

Its primary differentiator: document workflow generation. Say "write a project proposal for client X" and Artha produces a polished, formatted DOCX. Say "build a 12-month financial model" and it produces a working XLSX. No cloud. No subscriptions. Your data stays yours.

---

✨ Features

Feature	Status
🤖 Agentic ReAct loop (plan → execute → self-correct)	🚧 In development
📄 DOCX generation from natural language	🚧 In development
📊 PPTX presentation generation	🚧 In development
📈 XLSX spreadsheet generation	🚧 In development
📑 PDF report generation	🚧 In development
🔌 MCP-native tool system (any MCP server = a skill)	🚧 In development
🧠 Local RAG over your files (Ollama embeddings)	🚧 In development
🦙 Ollama first-class (auto-detect, pull, switch models)	🚧 In development
🔒 Zero telemetry by default	✅ By design
🐳 Docker-sandboxed tool execution	🚧 In development

---

🏗️ Built on

Artha is distributed as proprietary software, but it incorporates and builds upon open-source work, used under the MIT License — see NOTICE for the preserved upstream copyright and license text:

• OpenCoworkAI/open-cowork (MIT) — source code incorporated and modified under MIT; attribution preserved in NOTICE.
• OpenHands (MIT) — ReAct/CodeAct orchestration patterns (influence).
• Jan AI (MIT) — local LLM management UI/UX inspiration (influence).

Artha is also built with open-source npm dependencies, each under its own license — see THIRD-PARTY-NOTICES.md.

The original code Shree Labs Inc. added on top is proprietary (see LICENSE). The "Artha" name, the अ mark, and logo are trademarks of Shree Labs Inc. — see TRADEMARK.md.

---

🚀 Quick Start

Prerequisites

• Node.js 22+
• Ollama installed and running (ollama serve)
• At least one model pulled: ollama pull llama3.2:3b-instruct-q4_K_M
• Docker Desktop (optional — for sandboxed tool execution)

Development

# Clone
git clone https://github.com/artha-apps/artha.git
cd artha

# Install all workspace dependencies
npm install

# Start dev server (renderer + electron)
npm run dev

Build for distribution

npm run dist
# Output: dist/Artha-0.1.0.dmg  (macOS)
#         dist/Artha Setup 0.1.0.exe  (Windows)
#         dist/artha_0.1.0_amd64.deb  (Linux)

---

📁 Project Structure

artha/
├── packages/
│   ├── app/                  # Electron main process (Node.js + TypeScript)
│   │   └── src/
│   │       ├── main.ts       # Electron entry point
│   │       ├── preload.ts    # Secure IPC bridge (contextBridge)
│   │       ├── agent/        # ReAct orchestration engine
│   │       ├── llm/          # Single OpenAI-compat LLM client
│   │       ├── mcp/          # MCP tool registry
│   │       ├── rag/          # Local vector store + indexer
│   │       ├── docs/         # Document generation (DOCX/PPTX/XLSX/PDF)
│   │       ├── db/           # SQLite schema + migrations
│   │       └── ipc/          # IPC handler registration
│   └── renderer/             # React 19 UI (Vite + Tailwind + shadcn)
│       └── src/
│           ├── App.tsx
│           ├── stores/       # Zustand state
│           └── components/   # Chat, Sidebar, ExecutionLog, PlanApproval
├── docs/                     # Architecture docs
├── assets/                   # App icons
└── scripts/                  # Dev helpers

---

🧠 Architecture

Artha uses a TypeScript-first stack for v1 — one runtime across main process and renderer, zero cross-language IPC:

[React UI] ←— contextBridge IPC —→ [Node.js Main Process]
                                           │
                          ┌────────────────┼────────────────┐
                          ▼                ▼                ▼
                   [Agent Orchestrator] [MCP Registry] [Doc Generator]
                   (ReAct / CoT loop)  (MCP SDK)      (docx/pptx/xlsx)
                          │
                    [LLM Client]  ←— OpenAI-compat REST
                          │
                   [Ollama / LM Studio / llama.cpp]

See docs/architecture.md for full detail.

---

🤝 Contributing

Artha is proprietary and not currently open to outside code contributions. If you'd like to report a bug or request a feature, please open an issue. See CONTRIBUTING.md.

---

📄 License

Proprietary — Copyright © 2026 Shree Labs Inc. All rights reserved. See LICENSE. Third-party dependencies: see THIRD-PARTY-NOTICES.md.

---

_{Artha (अर्थ) — Sanskrit for purpose, meaning, livelihood. Your AI co-worker, locally.}