Open Source · Local-First · Privacy-First
Nekoni — open source local AI agent mascot logo

Your AI Agent,
On Your Machine

Run a private AI agent locally using Ollama. Access it from your phone via encrypted P2P. No subscriptions, no cloud, no compromise.

Ollama
Local inference
P2P WebRTC
Communication
Ed25519
Auth

Why Nekoni

Everything you need, nothing you don't

A complete local AI stack — inference, memory, connectivity, and automation — built for privacy-conscious developers.

Runs On Your Machine

All inference happens locally via Ollama. Your data never leaves your device. No monthly fees, no vendor lock-in.

P2P WebRTC Connectivity

Direct encrypted channel between your phone and agent. Only SDP/ICE exchange hits a public signal server — no traffic proxied.

Knowledge Base (RAG)

Upload PDFs, Office docs (DOCX, XLSX, PPTX), HTML, and images — OCR included via Docling. Indexed locally with ChromaDB and queryable from chat.

Skills & Automation

Define named prompt templates as skills. Schedule them with cron to run daily summaries, digests, or any recurring task automatically.

Cryptographic Auth

Ed25519 keypairs for both agent and mobile. QR code pairing on local network. Four-step mutual auth before any message is accepted.

Extensible Tools

Give the agent new capabilities by writing minimal Python. Drop a file in the tools directory and it becomes available immediately.

Getting Started

Up and running in minutes

One command. No cloud accounts, no API keys required.

01

Install & Run Locally

Run the install script. Everything installs automatically — Ollama pulls the model, the agent starts, and the web dashboard is ready.

$ install.sh
02

Pair Your Phone

Open the dashboard, scan the QR code with the Nekoni mobile app on your local network. Ed25519 pairing is done in seconds.

$ scan QR → paired
03

Chat, Upload, Automate

Chat with your agent anywhere. Upload documents for RAG. Create skills and schedule them. Everything stays on your machine.

$ it just works
terminal
$ curl -sSL https://raw.githubusercontent.com/nekonihq/nekoni/refs/heads/main/install.sh | bash
# Agent running at localhost:8000
Available on Web & iOS · Android Coming Soon

Your agent in your pocket

The Nekoni app connects to your home agent over an encrypted P2P channel. Chat, browse your knowledge base, run skills — from anywhere in the world with zero data passing through our servers.

  • Chat with your local LLM from anywhere
  • Manage knowledge base documents
  • Create and trigger skills on demand
Nekoni mobile app chat interface with local AI agent
Nekoni mobile app skills and automation panel
Nekoni mobile app settings and P2P connection configuration
Inference
OllamaLocal LLMsentence-transformers
Agent
Python / FastAPIChromaDBSQLite
Dashboard
React 19ViteRadix UI
Connectivity
WebRTCNode.js SignalEd25519
Web App
React PWAViteWebRTC DataChannel
Mobile
React NativeExpoWebRTC DataChannel

Architecture

A full stack that you control

Nekoni is a monorepo with five distinct apps — agent, signal server, dashboard, web app, and mobile — tied together with pnpm workspaces and a handful of Make targets.

The signal server only brokers WebRTC handshakes. Once connected, all communication goes through an end-to-end encrypted DataChannel directly between your phone and your machine.

pnpm workspaces
Monorepo tooling
Make
One-command workflow
ChromaDB
Local vector search
WebRTC
Direct P2P channel
Nekoni open source local AI agent logo

Own your AI. Own your data.

Star the repo, try it out, and help shape what personal AI computing looks like.