Skip to content

FullFran/nyxterm

Repository files navigation

nyxterm

AI-native terminal emulator. Tauri 2 + Rust. Tokyo Night. Pi embedded as the AI layer. Built for Nix homies.

status license phase

Quick start

# Prerequisites: Rust (1.70+), Node.js (18+), pnpm, libwebkit2gtk-4.1-dev (Linux)
pnpm install
pnpm tauri dev

The terminal opens a PTY session running your $SHELL. Type echo hello to verify the echo round-trip works.

Why

Today my terminal workflow is fragmented across:

  • Ghostty for the PTY surface
  • tmux for workspace multiplexing
  • opencode / claude-code for AI assistance
  • Chrome for YouTube and WhatsApp Web sitting next to my terminal

That's four moving parts to keep in sync. nyxterm collapses them into one cohesive surface: native multiplexing, AI nativa as a first-class panel, and (eventually) embedded web panes for the handful of sites I always have open alongside code.

Built around three principles:

  1. Concepts > Code. Get the architecture right before optimizing keybinds.
  2. Pi philosophy. Minimalism wins. 4 core tools (Read/Write/Edit/Bash), everything else is a skill.
  3. AI-first, human-led. The agent suggests; the human directs.

Stack

  • Backend: Rust + portable-pty + Tauri 2
  • Frontend: React + TypeScript + xterm.js (with fit, search, webgl, web-links, serialize addons)
  • AI layer: Vercel AI SDK v6+ with Anthropic / OpenAI / Ollama / MLX
  • Memory: Engram embedded, with MCP fallback
  • Packaging: Nix flake (Linux first), AppImage, then macOS + Windows

Roadmap

See ROADMAP.md for the full phase breakdown and the Issues page for live epics.

Phase Focus Milestone
1 MVP terminal puro (PTY, multiplexing, theme, shell integration) Phase 1 — MVP
1.5 Productivity (fuzzy picker, atuin palette, inline suggestions, project switcher) Phase 1.5
2 Embed Pi (pi-agent-core, pi-ai, pi-coding-agent), Engram first-class, MCP strategy decision Phase 2
3 Web panes — generic webview, libmpv for media, WhatsApp Web pane Phase 3

Inspirations

Contributing

See CONTRIBUTING.md. The repo follows Spec-Driven Development (SDD): proposal → spec → design → tasks → apply → verify → archive.

Agent context is in AGENTS.md. Skills live under .agents/skills/.

License

MIT — see LICENSE.

About

AI-native terminal emulator. Tauri 2 + Rust + xterm.js. Tokyo Night. Pi-style AI harness.

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages