Open source · Local-first · MIT

Serious work. Fully local.

Artha is a local-first AI workspace that generates documents, indexes your files, and runs agents — all on your hardware. No cloud calls. No accounts. No telemetry.

Loading latest release…
On-device execution
100%
Telemetry events
0
Platforms
3
License
MIT

An AI workspace that earns trust.

Every capability built around a single rule: your files, prompts, embeddings, and model outputs never leave the machine they were generated on.

01 / DOCUMENTS

Native artifacts

Generate Word, Excel, PowerPoint, and PDF files from natural language. Every artifact stays on your machine.

02 / TOOLS

MCP-native

Any Model Context Protocol server becomes a skill your agent can use. Bring your own integrations.

03 / RETRIEVAL

Local RAG

Index folders with Ollama embeddings. Search, cite, and chat with your files — no cloud.

04 / MODELS

Ollama-first

Auto-detect installed models, pull new ones, and switch between Llama, Mistral, Qwen, and more.

05 / PRIVACY

Zero telemetry

No accounts. No tracking. No background phone-home. The binary is auditable.

06 / SAFETY

Sandboxed tools

Optional Docker sandbox for untrusted tools. Run agents without giving them your shell.

Before your first launch.

Three things worth knowing.

  1. 01

    Requires Ollama running locally.

    We recommend llama3.2:3b-instruct-q4_K_M to start — fits in 8 GB of RAM and is fast on most laptops.

  2. 02

    Installers are currently unsigned.

    On macOS, right-click the app and choose Open. On Windows, click “More info” → “Run anyway”. Signed builds are on the roadmap.

  3. 03

    The binary is open source and auditable.

    Run it as a black box, or build from source — your call.