Skip to content

polyphon-ai/polyphon

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

320 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Polyphon

One chat. Many voices.

CI GitHub release License Follow on X

Polyphon screenshot

Polyphon is an Electron desktop application for orchestrating conversations between multiple AI agents simultaneously. Agents can respond to the user and to each other — like a conductor leading an ensemble.


Features

  • Multi-agent sessions — add any number of voices to a conversation and watch them reason together
  • Compositions — save and reuse named multi-voice configurations across sessions
  • Mixed providers — combine API-key voices (Anthropic, OpenAI, Gemini), CLI voices (claude, codex), and custom OpenAI-compatible endpoints (Ollama, LM Studio, vLLM) in the same session
  • Tones — per-voice tone overrides (professional, collaborative, concise, exploratory, teaching) or custom tones
  • Conductor profile — set your name, pronouns, and context so voices address you correctly
  • @mention routing — direct a message to a specific voice or broadcast to all
  • Local-first — no cloud dependency required; your data stays on your machine
  • No telemetry — never phones home without explicit opt-in

Voice Providers

Provider Type Requires
Anthropic (Claude) API ANTHROPIC_API_KEY
OpenAI API OPENAI_API_KEY
Google Gemini API GEMINI_API_KEY
Claude Code (claude) CLI claude CLI installed
Codex (codex) CLI codex CLI installed
Custom (Ollama, LM Studio, vLLM, …) OpenAI-compatible Endpoint URL

API keys are read from your shell environment — set them in .zshrc / .bash_profile and they are available even when launching from the Dock.


Installation

Download

Pre-built installers for macOS (Apple Silicon) are available on the Releases page.

Build from source

Prerequisites: Node.js 22+, npm

git clone https://github.com/polyphon-ai/polyphon.git
cd polyphon
make install
make run

Development

make install        # install dependencies + git hooks
make run            # start in development mode (hot reload)
make build          # package the Electron app
make dist           # create distributable installers

Testing

make test              # unit + integration + e2e
make test-unit         # Vitest unit tests only
make test-integration  # Vitest integration tests only
make test-e2e          # Playwright e2e with mocked voices (CI-safe, no credentials needed)
make test-e2e-live     # e2e against real providers (opt-in, never CI)
make test-openai-compatible-live # e2e against real openai compatible providers (ollama)
make test-watch        # Vitest in watch mode

Lint

make lint           # TypeScript type-check (no emit)

Developer tools

Chrome DevTools are closed by default in make run. To open them at launch:

POLYPHON_DEVTOOLS=1 make run

Community


Contributing

Issues and bug reports are welcome. This project is not currently accepting pull requests.


License

Apache 2.0 © 2026 Corey Daley


Tech Stack

Layer Technology
Shell Electron 41
UI React 19 + TypeScript + Tailwind CSS v4
State Zustand 5
Database better-sqlite3 v12 + SQLCipher 4.14.0 (AES-256 whole-database encryption)
Build Electron Forge + Vite 7
Testing Vitest 4 + Playwright

About

Open source Electron desktop app for orchestrating conversations between multiple AI agents simultaneously. One chat. Many minds.

Topics

Resources

License

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Sponsor this project

 

Packages

 
 
 

Contributors