Your terminal as an AI agent control room

What it does

Snow CLI is a terminal application that orchestrates AI coding agents across multiple LLM providers. It wraps conversation, code generation, and task execution in an Ink-based text interface, and maintains local profiles, sessions, and async tasks in a ~/.snow/ directory. The project also ships companion IDE extensions for VSCode and JetBrains.

The interesting bit

Most AI CLI tools stream text and stop; Snow CLI treats the terminal as an application platform in its own right. It offers sub-agents, workflow hooks, LSP integration, MCP services, custom search plugins, and a headless server mode — effectively turning a Node.js terminal UI into a configurable agent host.

Key highlights

• Supports multiple LLM backends through API adapters (OpenAI, Gemini, Claude, and Deepseek).
• Extensive customization surface: custom hooks, sub-agents, status-line plugins, search engines, and themes.
• Team mode for multi-agent collaboration and parallel task execution.
• Headless and SSE service modes for scripting and third-party tool integration.
• Bundles IDE extensions for VSCode and JetBrains alongside the terminal core.

Caveats

• The README acts mostly as a documentation index, listing many feature guides but offering almost no demonstration of the core agent loop or UI in action.
• With sub-agents, hooks, MCP, LSP, and custom plugins all competing for attention, the day-to-day workflow can be hard to visualize from the sources alone.

Verdict

Developers who want a highly configurable, terminal-native AI workbench with IDE extensions and multi-agent support should look here. Those seeking a simple, drop-in coding assistant may find the configuration surface overwhelming.