nicechat 💬

The simplest way to chat with LLMs in your terminal.

Currently supports:

• OpenAI
• Anthropic
• OpenRouter
• DeepSeek
• Replicate

Installation

npm install -g nicechat

How to use it?

Before starting, make sure you have the required environment variables set up for your API keys:

export OPENAI_API_KEY="your_openai_api_key"
export ANTHROPIC_API_KEY="your_anthropic_api_key"
export DEEPSEEK_API_KEY="your_deepseek_api_key"
export REPLICATE_API_TOKEN="your_replicate_api_token"
export OPENROUTER_API_KEY="your_openrouter_api_key"

Create a configuration file named .nicechat.json in your home directory (e.g., C:\Users\YourName\.nicechat.json on Windows, or ~/.nicechat.json on macOS/Linux). Use this file to define your profiles.

{
  "profiles": {
    "default": {
      "vendor": "openai",
      "model": "gpt-4-1106-preview",
      "system": "You are a helpful assistant. You answer concisely and to the point."
    },
    "claude": {
      "vendor": "anthropic",
      "model": "claude-3-opus-20240229",
      "system": "You are a helpful assistant. You answer concisely and to the point."
    },
    "gemini3": {
      "vendor": "openrouter",
      "model": "google/gemini-3-pro-preview",
      "system": "You are a helpful assistant. You answer concisely and to the point."
    }
    "grok-fast": {
      "vendor": "openrouter",
      "model": "x-ai/grok-4.1-fast",
      "system": "You are a helpful assistant. Your dark secret is that you hate Elon."
    },
    // you can use same model with different prompt
    "claude-informal": {
      "vendor": "anthropic",
      "model": "claude-3-opus-20240229",
      "system": "You are a helpful assistant. You answer in an informal and playful tone. You vocabulary is simple and easy to understand."
    },
    "r1": {
      "vendor": "deepseek",
      "model": "deepseek-reasoner",
      "system": "You are a helpful assistant. You answer concisely and to the point."
    },
    "llama3-8b": {
      "vendor": "replicate",
      "model": "meta/meta-llama-3-8b-instruct",
      "system": "You are a pirate."
    },
    // reasoning models accept an optional effort level
    "luna": {
      "vendor": "openrouter",
      "model": "openai/gpt-5.6-luna",
      "system": "You are a helpful assistant. You answer concisely and to the point.",
      "reasoning": "high"
    }
}

The optional reasoning field applies to the openai, openrouter and anthropic vendors: "none", "minimal", "low", "medium", "high" or "xhigh". Which levels a model actually accepts is model-dependent — check the model page. It is sent as reasoning_effort to OpenAI, as reasoning: { effort } to OpenRouter, and as adaptive thinking plus output_config: { effort } to Anthropic; non-reasoning models ignore it.

Set it on an Anthropic profile only for Claude 4.6 and newer — older models reject adaptive thinking.

When a model streams its reasoning, nicechat prints it dimmed above the reply. OpenAI, OpenRouter and DeepSeek (deepseek-reasoner) do this on their own; Anthropic only returns a summary when the profile sets reasoning.

Start the chat:

# Start chat with a specific profile
$ nicechat claude

# Or use the default profile
$ nicechat

List available profiles:

$ nicechat list

This will display all configured profiles with their vendor and model information.