nekoni
Health Warn
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 5 GitHub stars
Code Pass
- Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
- Permissions — No dangerous permissions requested
This project is a self-hosted AI agent that runs entirely on your local machine. It allows you to privately interact with local LLMs, RAG, and automated skills directly from your phone using peer-to-peer WebRTC, keeping all data on your hardware.
Security Assessment
Risk: Medium. The tool enables local LLM inference and allows users to build extensible tools using Python, which inherently carries code execution risks depending on how it is configured. It requires network access to facilitate P2P WebRTC connections and uses a public signaling server to establish the link. While it features Ed25519 key-pair authentication for mutual security, the installation method relies on piping a remote shell script directly into bash (`curl | bash`). A light code scan of 12 files found no dangerous patterns or hardcoded secrets, and it does not request dangerous system permissions.
Quality Assessment
The project is actively maintained with recent pushes and is properly licensed under MIT. However, it currently has extremely low community visibility with only 5 GitHub stars. This means it has not undergone widespread peer review or battle-testing by a larger developer community.
Verdict
Use with caution: the code is currently clean, but the low community adoption and remote bash installation method mean you should review the setup script and custom skills yourself before deploying on your machine.
Local AI agent on your home machine, accessible from your phone via P2P WebRTC. Ollama + RAG + Skills + cron - no cloud, no subscription, your data stays on your hardware.
🐱 Nekoni — Local AI Agent You Control From Your Phone
nekoni.dev · app.nekoni.dev · iOS App
Run your own AI. Own your data. Access it from your phone — no cloud required.
Nekoni is a self-hosted AI agent that runs on your machine and connects directly to your phone via WebRTC.
No cloud. No subscriptions. No data leaving your hardware.
Features
- Fully local — LLM inference via Ollama, embeddings via sentence-transformers, vector search via ChromaDB
- Phone access — connect from mobile or web over direct WebRTC DataChannel
- No cloud relay for data — public signaling is used only for SDP/ICE exchange
- Key-pair security — Ed25519 identity keys and mutual authentication
- RAG — ingest documents and query them from chat
- Skills — reusable prompt templates with cron scheduling
- Extensible — add tools in a few lines of Python
- Observable — live trace stream in the dashboard
Quick Start
curl -sSL https://raw.githubusercontent.com/nekonihq/nekoni/refs/heads/main/install.sh | bash
Installs all dependencies, pulls the default model, and starts Nekoni. Requires ~4 GB disk space for the model.
Stop / start:
cd ~/.nekoni && make down
cd ~/.nekoni && make up
Requirements
- macOS or Linux (Windows: use WSL2)
- Docker Desktop — installed automatically (Homebrew on macOS, get.docker.com on Linux)
- ~4 GB disk space for the Ollama model
How It Works
flowchart TD
CLIENT["📱 Phone / Browser"]
WEBRTC["🔗 WebRTC (P2P)"]
AGENT["🏠 Your Machine (Nekoni)"]
MODEL["🧠 Local Model (Ollama)"]
CLIENT --> WEBRTC --> AGENT --> MODEL
Nekoni uses direct peer-to-peer communication between your device and your home machine. The public signal server is only used to establish the WebRTC connection.
More details:
Use Cases
- Private AI assistant at home
- Local alternative to cloud AI tools
- AI-powered automations on your own hardware
- Experimenting with local-first AI workflows
- Building custom tools and agent skills
Documentation
⭐ Support
If you find Nekoni useful:
- Star the repo
- Share your setup
- Build something on top of it
- Open issues and contribute improvements
License
MIT
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found