Introduction

Eigin is the application layer for personal AI: a persistent agent that runs on your device, connects any AI inference provider you want, and builds knowledge about you across every conversation.

These are the technical docs for Eigin: what it is, how it works, and how its implemented.

Features

  • Persistent agent. One per device, with its own identity, knowledge graph, intents, and skills. See The Agent.
  • Local-first. Chats, knowledge, intents, and settings live on the device. See System context.
  • Model-agnostic. Cloud (any OpenAI-compatible API) or fully local. See Inference providers.
  • Voice. On-device dictation and a live call mode for hands-free conversation. See Voice.
  • Free and open source. See Philosophy.

Where to start

If you're new to the codebase, read in this order:

  1. Philosophy — why Eigin exists.
  2. System context — Eigin at the system level.
  3. The Agent — the agent ontology: Core, Experience, Knowledge, Intents, Skills.
  4. iOS Implementation — how the iOS app realises those concepts.

Then jump into whatever you're touching: Conversation, Tools, or Voice.

Building from source

This is a monorepo containing all Eigin applications. Install dependencies for the platform you want to work on.

iOS:

  • Xcode — xcode-select -s /Applications/Xcode.app/Contents/Developer
  • XcodeGen — brew install xcodegen
  • SwiftFormat — brew install swiftformat

Relay:

Command runner (required for all platforms):

  • just — brew install just

After cloning, run just hooks once to install git hooks. Then use just build, just test, and just format, scoped by platform if needed:

just build          # All targets
just build ios      # iOS only
just build relay    # Relay only

Run just --list to see all available commands.