OpenJarvis — Personal AI OS

Your personal AI that learns while you sleep, answers from your own knowledge, and executes tasks with AI agents.

What It Does

Every day, OpenJarvis automatically:

• Researches topics you care about from 9 sources (Wikipedia, arXiv, Reddit, GitHub, HN, Stack Overflow, dev.to, RSS, YouTube)
• Stores everything in your personal Obsidian vault
• Indexes all notes into a vector database
• Lets you ask questions answered from YOUR knowledge — not generic internet

Then when you need to act:

• Routes your task to the right AI skill automatically
• Executes with Ruflo multi-agent swarms
• Gets smarter every single day without any manual input

Install (60 seconds)

npx openjarvis init

That's it. Answer 3 questions. Your brain starts growing.

Requires:

• Python 3.10+
• Node.js 18+
• Anthropic API key (get one at console.anthropic.com — ~$1.20/month)

Usage

# Ask your brain anything
npx openjarvis ask "how does gradient descent work"
npx openjarvis ask "best practices for Python async"
npx openjarvis ask "what is neuroplasticity"

# Interactive mode
npx openjarvis ask

# Fetch knowledge now (runs automatically daily)
npx openjarvis run

# System status
npx openjarvis status

How It Works

npx openjarvis init
        ↓
BrainFeeder runs daily at 9 AM
        ↓
Researches 12 topics from 9 sources
        ↓
Creates 24 Obsidian notes with AI enrichment
        ↓
Vector indexer embeds all notes into ChromaDB
        ↓
jarvis.py searches your vault for any query
        ↓
Claude answers using YOUR notes as context
        ↓
Ruflo agents execute complex tasks

What Gets Created Daily

For each topic, OpenJarvis creates:

| File | Contents | |---|---| | Summary.md | TL;DR, 5 Brain Bullets, Flashcards, Reddit/HN/GitHub sources | | Deep Dive.md | Full Wikipedia breakdown, arXiv research papers, Concept Map | | Category MOC.md | Auto-updated Map of Content |

Example Output

$ npx openjarvis ask "how does gradient descent work"

  Searching your brain...
  Found 5 relevant notes
  Routing to: Thinking -> brainstorming

  # How Gradient Descent Works

  Based on your vault knowledge...

  Sources from your vault:
  - Gradient descent — Deep Dive (07-AI-Brain) [0.701]
  - Gradient descent — Summary (07-AI-Brain) [0.607]
  - PyTorch (Imported Skills) [0.411]

Topics Covered (235 total, auto-rotating)

• AI & Machine Learning — LLMs, Transformers, RAG, Agents, Diffusion Models
• Neuroscience & Memory — Neuroplasticity, Spaced Repetition, Flow State
• Programming — Python, System Design, Docker, FastAPI, Clean Code
• Hacking — Penetration Testing, OWASP, OSINT, CTF
• Automation — n8n, LangChain, GitHub Actions, Self-hosted
• Linux — Bash, SSH, Networking, Security Hardening
• Data Science — Pandas, Statistics, Time Series, Visualization
• Mental Models — First Principles, Stoicism, Decision Frameworks
• Biohacking — Sleep, Nootropics, HRV, Longevity

Add your own topics in config.json — any topic, any category.

Architecture

OpenJarvis/
├── brain_feeder.py        Daily knowledge intake (9 sources)
├── jarvis.py              CLI — answer + execute
├── config.json            Your topics + settings
├── scripts/superbrain/
│   ├── vector_indexer.py  Embeds notes to ChromaDB
│   ├── brain_query.py     Semantic vault search
│   └── vault_mcp_server.py MCP server for Claude Code

Claude Code Integration

OpenJarvis exposes your vault as MCP tools in Claude Code:

search_brain(query)     Semantic search all your notes
route_task(task)        Layer 3 hub + skill routing
get_note(title)         Fetch any note by title
brain_context(query)    Get vault context for agents
list_categories()       Show vault structure

After npx openjarvis init, restart Claude Code — your brain is available in every session.

Ruflo Integration

Complex tasks get routed to Ruflo multi-agent swarms:

npx openjarvis ask --mode agent "set up a CI/CD pipeline for my FastAPI project"

Jarvis searches your vault for context, enriches the task with your knowledge, then hands it to the right Ruflo specialist agent.

Cost

| | | |---|---| | Anthropic API | ~$0.02/day (~$0.60/month) | | Ruflo | Free (open source) | | Everything else | Free |

The $5 free credit from Anthropic lasts ~8 months at this usage rate.

Roadmap

• [ ] Web UI dashboard
• [ ] Topic marketplace (share topic lists)
• [ ] Team vaults (shared knowledge)
• [ ] Mobile app
• [ ] More sources (Podcast transcripts, Papers With Code)

Built With

• BrainFeeder — Knowledge intake engine
• ChromaDB — Local vector database
• sentence-transformers — Embeddings
• Ruflo — Multi-agent orchestration
• Anthropic Claude — AI backbone

License

MIT — free forever, self-hosted, your data stays on your machine.

Built by @Suprithg95