A complete CLI Agent for production and enterprise level, with incredible results.

GitHub

• CLI + TUI — full terminal UI with keyboard navigation, token display, improved history & compaction, overlay system
• Multi-provider — Zen (free, no config), Gemini, Qwen (DashScope), HuggingFace, skill search, custom providers (any OpenAI-compatible API)
• Skills system — folder-based skills with improved prompt integration and YAML frontmatter
• Tool execution — read/write files, run commands, search code, browse web, glob patterns
• Session management — real conversation history, persistent sessions with better compaction, full transcript replay, resume, export
• Context tracking — token estimator with per-model context limits, auto-compaction at 85%, updated to reflect new history behavior
• Multi-platform — embeddable in WhatsApp (Baileys), Discord, and Telegram bots, Although this function is not complete (It is under development)
• Background workers — detach long-running turns to background processes
• i18n — English and Spanish interfaces

npm install -g zyn-ai
zyn

git clone https://github.com/SoyMaycol/Zyn.git
cd Zyn
npm install
npm start

zyn                  # Interactive TUI (default)
zyn "question"       # Single prompt (CLI mode)
zyn --new            # Force new session
zyn --resume ID      # Resume existing session (keeps history)

14 built-in models across 4 providers. Each model has its real context length in the StatusBar (e.g. 10.8K/128K).

Default model: nemotron (Zen, no configuration required).

Press ESC twice in the TUI to stop the current task.

Add any OpenAI-compatible API:

# Interactive: run /providers → select "+ Add custom provider"
# Or manual:
/provider set groq baseUrl https://api.groq.com/openai/v1
/provider set groq apiKey gsk_xxxx
/provider set groq contextLength 128000
/provider sync groq

Configurable fields: apiKey, baseUrl, modelId, contextLength, email, password, modelEndpoint, chatEndpoint.

Skills are folders under data/skills/<name>/SKILL.md with YAML frontmatter:

---
name: reasoning
description: Reasoning and planning for complex tasks
---

# Skill body here

The system prompt automatically advertises every loaded skill to the model, and mentions of GitHub repos/skills trigger web_search to find them.

const { createAgent } = require('zyn-ai');

const agent = createAgent({
  model: 'nemotron',
  language: 'en',
  autoApprove: false,
});

const response = await agent.send('userId', 'Hello!');

const { createAgent, platforms } = require('zyn-ai');

const agent = createAgent({ model: 'nemotron' });

// WhatsApp (Baileys)
await platforms.whatsapp({ agent, session: './whatsapp-auth' });

// Discord
await platforms.discord({ agent, token: 'DISCORD_TOKEN' });

// Telegram
await platforms.telegram({ agent, token: 'TELEGRAM_TOKEN' });

Install the corresponding optionalDependencies only if you need them.

MIT - Maycol B.T