JARVIS is an autonomous AI daemon designed for persistent, 24/7 operation, offering desktop awareness and multi-machine control. It targets power users and engineers seeking to automate complex workflows and achieve goals through an AI agent that can see, act, and delegate across their digital environment. The system provides an always-on assistant that integrates deeply with user desktops and servers, enabling proactive assistance and structured task management.

How It Works

JARVIS employs a distributed architecture consisting of a central daemon and lightweight sidecar agents. The daemon acts as the core intelligence, managing LLM connections, memory, and decision-making, runnable on a server or local machine. Sidecars, implemented as Go binaries, are installed on individual machines (laptops, desktops) to grant the daemon access to local resources like the screen, applications, terminal, and filesystem. This separation allows the daemon to remain accessible while orchestrating actions across multiple machines, leveraging a multi-agent hierarchy with specialist roles and a visual workflow builder for complex task execution.

Quick Start & Requirements

Install the daemon globally via Bun: bun install -g @usejarvis/brain, followed by jarvis onboard for interactive setup. A Docker installation is also available. Prerequisites include Bun >= 1.0, at least one LLM API key (Anthropic, OpenAI, Gemini, or Ollama), and optionally Google OAuth, Telegram/Discord tokens, or an ElevenLabs API key. Native daemon installation supports macOS, Linux, and WSL; Windows users must use WSL2 or Docker. The dashboard is accessible at http://localhost:3142.

Highlighted Details

• Always-On Daemon & Cross-Machine Reach: Operates 24/7, connecting to unlimited sidecars for comprehensive desktop, browser, terminal, and filesystem control across any machine.
• Multi-Agent Hierarchy & Visual Workflows: Features 9 specialist agent roles and a visual workflow builder with over 50 node types for complex automation.
• Desktop Awareness & Native Control: Utilizes screen capture (5-10s intervals) with hybrid OCR and Go-based sidecars for native application automation on Windows, Linux, and macOS.
• Goal Pursuit (OKR): Implements an OKR framework for defining objectives, tracking key results, and providing daily planning, evening reviews, and escalation for accountability.

Maintenance & Community

The project is built using Bun (runtime), TypeScript (daemon/UI), Go (sidecar), React 19, and Tailwind CSS 4. Development is active, with 16 milestones completed and a detailed roadmap available in VISION.md. Community interaction is facilitated via Discord and GitHub Issues.

Licensing & Compatibility

JARVIS is licensed under the Jarvis Source Available License 2.0 (based on RSALv2). This license is source-available but not a standard OSI-approved open-source license, potentially imposing restrictions on commercial use, modification, and distribution.

Limitations & Caveats

Native daemon installation is not supported on Windows; WSL2 or Docker must be used. Dockerized deployments are isolated and require sidecars for host machine interaction. The RSALv2 license may present compatibility challenges for certain commercial or closed-source integration scenarios.