Aug 2025 - Present
Jarvis
Closed-source, CLI-based agentic AI program written in Python.
Overview
Jarvis is an agentic-based artificial intelligence program written in Python. It is powered by third-party LLMs including Claude Opus 4.6 and Gemini 3.1 Pro, as well as self-hosted and fine-tuned DeepSeek full models and distillates.
Jarvis is CLI-based and runs on Windows, macOS, and Linux (with support for Debian-based and RHEL-based distributions).
Its core execution layer is built around self-hosted full models and fine-tuned distillates, while complex tasks are offloaded to larger enterprise models such as Claude Opus 4.6 and Gemini 3.1 Pro.
It is designed to perform virtually any computer task, including software operations, browser control, infrastructure access, and autonomous task execution.
Core Capabilities
- Open and control desktop applications.
- Navigate the web by controlling a web browser directly.
- Use other LLM systems efficiently, including tools like Perplexity.
- SSH into servers and execute commands.
- Create, organize, and modify files/directories across local project workspaces.
- Run developer tooling such as tests, builds, linters, and package operations.
- Inspect logs, monitor running processes, and assist with debugging system or app issues.
- Draft technical documentation, release notes, and implementation summaries from project context.
- Automate repeatable CLI sequences as reusable command-driven task chains.
- Develop full-scale software directly, or orchestrate other, more specialized AI systems (including OpenAI Codex) to improve efficiency and execution quality.
- Read recent emails and draft or send replies autonomously in a human-like style.
- Execute most other general computer tasks beyond the specific capabilities listed above.
Project Status
- Started development in August 2025 and remains in active development today.
- Closed source.
