Overview

Selene is an open-source desktop application for building, running, and deploying AI agents locally on your own machine rather than through cloud-hosted services. It ships with a multi-agent architecture that supports role-based agents, plus a powerful tool kit out of the box: browser automation via Chromium/Playwright, web search and scraping, Git integration with pull-request management, video generation via Remotion, a local LanceDB vector database, and a voice mode backed by 3D talking-head avatars. Agents can be deployed straight to Slack, Telegram, Discord, or WhatsApp channels. Selene supports a wide range of LLM backends — Anthropic Claude, OpenAI, Google Gemini, Minimax, Kimi, and 200+ models through OpenRouter, plus local models via Ollama. Uses a bring-your-own-API-key model. MIT licensed, runs on macOS (Apple Silicon and Intel) and Windows.

Pricing: Free, open source (BYOK for cloud models)

Architecture: Apple Silicon, Intel

Key Features

Multi-agent architecture with role-based agents
Local execution — no cloud lock-in
Browser automation via Chromium/Playwright
Web search and scraping
Voice mode with 3D talking-head avatars
Git integration with PR management
Video generation via Remotion
Local vector database (LanceDB)
Channel deployment to Slack, Telegram, Discord, and WhatsApp
Supports Anthropic, OpenAI, Gemini, Minimax, Kimi, Ollama, and 200+ models via OpenRouter
MIT licensed open source