Quick start

Command-line interface that enables interactive chat with LLMs.

Quick start

# Install binary
$ bash -c "$(curl -fsSL https://raw.githubusercontent.com/u1and0/gpt-cli/refs/heads/main/install.sh)"
$ sudo ln -s ./release-package/gpt /usr/bin

# Setup API key
$ export OPENAI_API_KEY='sk-*****'

# Run
$ gpt
Ctrl-D to confirm input, q or exit to end chat
You: hi
...
gpt-4o-mini: Hello! How can I assist you today?
You:

However, the API key must be set to an environment variable in advance.

Single mode (Not chat mode)

$ gpt -n "hi"
Hello! How can I assist you today?

Use pipeline

$ echo "hi" | gpt
Hello! How can I assist you today?

Forced output in single mode. (It does not shift to conversation mode.)

Command Integration Examples

For practical examples of how to integrate gpt-cli with other command-line tools, see the example.md file. This demonstrates piping data from tools like ddgr, w3m, jq, and others into gpt-cli for enhanced analysis and processing.

Installation

You have 3 options.

Download binary

The simplest way.

$ bash -c "$(curl -fsSL https://raw.githubusercontent.com/u1and0/gpt-cli/refs/heads/main/install.sh)"
$ sudo ln -s ./release-package/gpt /usr/bin
$ gpt -v

Almost same command here.

$ curl -LO https://github.com/u1and0/gpt-cli/releases/download/v0.9.5/gpt-cli-linux-x64.zip
$ unzip gpt-cli-linux-x64.zip
$ sudo ln -s ./release-package/gpt /usr/bin
$ gpt -v

gpt-cli-macos.zip and gpt-cli-windows.zip are also available. Try to switch zip file name.

Use runtime

Required deno

Create link on your path like ~/.deno/bin using deno install.

$ git clone https://github.com/u1and0/gpt-cli
$ cd gpt-cli
$ deno install -f --allow-net --allow-env --allow-read --name gpt gpt-cli.ts
$ export PATH=$PATH:~/.deno/bin
$ bash -l

Compile from source

Required deno

Create binary by deno compile.

$ git clone https://github.com/u1and0/gpt-cli
$ cd gpt-cli
$ deno compile --allow-net --allow-env --allow-read --output gpt gpt-cli.ts
$ sudo ln -s ./release-package/gpt /usr/bin

Setup

API keys

OpenAI API (for GPT)

Get OpenAI API key, then set environment argument.

export OPENAI_API_KEY='sk-*****'

Anthropic API (for Claude)

Get Anthropic API key, then set environment argument.

export ANTHROPIC_API_KEY='sk-ant-*****'

Google API (for Gemini)

Get Google API key, then set environment argument.

export GOOGLE_API_KEY='*****'

XAI API

Get XAI API key, then set environment argument.

export XAI_API_KEY='*****'

Groq API

Get Groq API key, then set environment argument.

export GROQ_API_KEY='*****'

Together AI API

Get Together AI API key, then set environment argument.

export TOGETHER_AI_API_KEY='*****'

Fireworks API

Get Fireworks API key, then set environment argument.

export FIREWORKS_API_KEY='*****'

MistralAI API

Get MistralAI API key, then set environment argument.

export MISTRAL_API_KEY='*****'

Replicate API

Get Replicate API token, then set environment argument.

export REPLICATE_API_TOKEN='*****'

Setup Ollama (for Local running Models)

Setup Ollama, see github.com/ollama/ollama
Download ollama model such as ollama pull modelName
Start ollama ollama serve on your server.

Usage

$ gpt -m gpt-4o-mini -x 8192 -t 1.0 [OPTIONS] PROMPT

Options

short option	long option	type	description
-v	--version	boolean	Show version
-h	--help	boolean	Show this message
-m	--model	string	LLM model (default gpt-4o-mini)
-x	--max_tokens	number	Number of AI answer tokens (default 8192)
-t	--temperature	number	Higher number means more creative answers, lower number means more exact answers (default 1.0)
-u	--url	string	URL and port number for ollama server
-s	--system-prompt	string	The first instruction given to guide the AI model's response.
-f	--file	string	Attachment file path. The Glob pattern can be used. (Such as `./test/*.ts`)
-n	--no-chat	boolean	No chat mode. Just one time question and answer.

PROMPT

A Questions for Model. Using regular expressions to match and generate the appropriate LLM instance allows you to use new models in 0 days after a new API is released.

Models

OpenAI
- gpt-4o-mini
- gpt-4o
- o1-mini
- o1
- o3-mini...
Anthropic
- claude-3-7-sonnet-latest
- claude-3-5-sonnet-20241022
- claude-3-5-sonnet-latest
- claude-3-opus-20240229
- claude-3-haiku-20240307
Gemini
- gemini-2.5-pro-exp-03-25
- gemini-2.0-flash...
- gemini-2.0-flash-thinking-exp...
Gemma
- gemma-3-27b-it...
Grok
- grok-3-latest
- grok-2-latest
- grok-2-1212
Groq
- groq/llama3.1-70b-specdec
- groq/llama-3.3-70b-specdec
- groq/deepseek-r1-distill-qwen-32b
- groq/deepseek-r1-distill-llama-70b
TogetherAI
- togetherai/deepseek-ai/DeepSeek-R1
- togetherai/deepseek-ai/DeepSeek-R1-Distill-Llama-70B-free
- togetherai/meta-llama/Llama-3.3-70B-Instruct-Turbo-Free
- togetherai/Qwen/QwQ-32B-Preview
- togetherai/meta-llama/Llama-3.1-405B-Instruct-Turbo
- togetherai/google/gemma-2-27b-it
- togetherai/mistralai/Mistral-7B-Instruct-v0.3...
Fireworks
- fireworks/accounts/fireworks/models/deepseek-r1
- fireworks/accounts/fireworks/models/llama-v3p1-405b-instruct
- fireworks/accounts/fireworks/models/deepseek-v3
MistralAI
- mistralai/codestral-latest
- mistralai/mistral-large-latest
- mistralai/mistral-small-latest
Replicate
- replicate/deepseek-ai/deepseek-r1
- replicate/meta/meta-llama-3-70b-instruct
- replicate/mistralai/mixtral-8x7b-instruct-v0.1
- replicate/snowflake/snowflake-arctic-instruct
- replicate/replicate/flan-t5-xl...
Ollama Use before $ ollama serve locally
- ollama/phi4
- ollama/llama3.3:70b
- ollama/mixtral:8x7b-text-v0.1-q5_K_M...

/ command

Help (/commands):

/?, /help Help for a command
/clear Clear session context
/modelStack Show model's history
/bye,/exit,/quit Exit

@ command

Help (@commands): Change model while asking.

@ModelName Change LLM model
ex) @gemini-1.5-pro your question...

Test

$ deno test --allow-env --allow-read --allow-write

See gpt-cli/test

Vim plugin (Optional)

This is a Vimmer-only option. This option brings a Github Copilot-like experience to your Vim.

For example, if you want to manage plug-ins with dein, write a toml file like this.

[[plugins]]
repo = 'u1and0/gpt-cli'
if = '''executable('gpt')'''
hook_add = '''
    let s:model = 'gemini-2.0-flash' "'claude-3-5-haiku-20241022'
    let s:code_model = 'claude-3-5-haiku-latest' " 'claude-3-5-sonnet-20241022'
    let s:pro_model = 'groq/deepseek-r1-distill-qwen-32b' " 'groq/deepseek-r1-distill-llama-70b' 'claude-3-5-sonnet-20241022'  gemini-2.0-flash
    let s:llama = 'groq/llama-3.3-70b-versatile'
    let s:url = 'http://localhost:11434'

    command! -nargs=0 -range GPTGenerateCode <line1>,<line2>call gptcli#GPT('You are best of code generator. Generate a prompt to continue coding based on the given input code. Generate only code effectively, DO NOT generate descriptions nor code blocks. If you need describe code, please comment out it.', { 'max_tokens': 500, 'temperature': 0.5, 'model': s:llama})
    " Keybind C-X, C-G
    inoremap <C-x><C-g> <Esc>:.-5,.+5GPTGenerateCode<CR>
    " Docs to code
    command! -nargs=0 -range GPTGenerateDocs <line1>,<line2>call gptcli#GPT('あなたの仕事は、提供されたコード スニペットを取り出し、シンプルでわかりやすい言葉で説明することです。コードの機能、目的、主要コンポーネントを分析します。最小限のコーディング知識を持つ人でも説明が理解できるように、例え話、例、わかりやすい用語を使用します。どうしても必要な場合を除き、専門用語の使用を避け、使用される専門用語については明確な説明を提供します。目標は、コードが何をするのか、そしてそれが高レベルでどのように機能するのかを読者が理解できるようにすることです。', {"max_tokens": 2000, "model": s:model})
    " Create test code
    command! -nargs=0 -range GPTGenerateTest <line1>,<line2>call gptcli#GPT('You are the best code tester. Please write test code that covers all cases to try the given code.', { "temperature": 0.5, "model": s:code_model})
    command! -nargs=0 -range GPTErrorBustor <line1>,<line2>call gptcli#GPT('Your task is to analyze the provided code snippet, identify any bugs or errors present, and provide a corrected version of the code that resolves these issues. Explain the problems you found in the original code and how your fixes address them. The corrected code should be functional, efficient, and adhere to best practices in programming.', {"temperature": 0.5, "model": s:code_model})
    command! -nargs=0 -range GPTCodeOptimizer <line1>,<line2>call gptcli#GPT("Your task is to analyze the code snippet and suggest improvements to optimize its performance. Identify areas where the code can be made more efficient, faster, or less resource-intensive. Provide specific suggestions for optimization, along with explanations of how these changes can enhance the code’s performance. The optimized code should maintain the same functionality as the original code while demonstrating improved efficiency.", { "model": s:code_model})
    command! -nargs=0 -range GPTShaper <line1>,<line2>call gptcli#GPT("Format the code without rewriting the content of the code.", { "model": "gemini-2.0-flash-lite-preview-0205", "temperature": 0})

    " Any system prompt
    command! -nargs=? -range GPTComplete <line1>,<line2>call gptcli#GPT(<q-args>, { 'model': s:model})
    " command! -nargs=? -range GPTComplete <line1>,<line2>call gptcli#GPT(<f-args>, {'model': 'llama3.2:1b', 'url': 'http://localhost:11434' })

    " Conversate with GPT
    command! -nargs=? GPTChat call gptcli#GPTWindow(<q-args>, {'model': s:model})
    command! -nargs=? GPTProChat call gptcli#GPTWindow(<q-args>, {'model': s:pro_model})
    command! -nargs=? GPTWizard call gptcli#GPTWindow(<q-args>, 'あなたは偉大な老魔術師であり、Linuxエンジニアです。コマンドを魔術に見立てて、儀式を弟子に伝授する形式で質問に答えるのじゃ。' , {'model': s:llama, 'temperature': 1.04})

    " Conversate with file content
    command! -nargs=* GPTChatWithFile call gptcli#GPTWindow(<f-args>, {'model': s:model})
'''

By setting your own prompt in the GPT() function, the AI will return a one-shot document based on your own document with your own arrangements.

The GPTWindow() function allows you to open a terminal and interact with AI on Vim by putting a case-by-case system prompt in the command line.

See more example

LLM Terminal Clients Comparison

This section compares gpt-cli with other popular LLM terminal clients to help you understand the similarities and differences between these tools.

Common LLM Terminal Clients

Terminal-Based LLM Clients

Client Name	Description
`gpt-cli`	Command-line interface for interactive chat with various LLMs including OpenAI, Anthropic, Google, Groq, and local models.
`ollama`	Tool for running large language models locally with a simple API.
`Claude Code`	Anthropic's command-line tool for delegating coding tasks to Claude directly from the terminal.
`llm`	CLI tool by Simon Willison for interacting with LLMs, with plugin support.
`shell\_gpt` (sgpt)	Command-line interface to chat with ChatGPT from the terminal.
`aichat`	Chat with AI models in your terminal with support for multiple providers.
`lm-terminal`	Minimalist terminal interface for interacting with LLMs.
`openai-cli`	Official OpenAI command-line interface for their models.
`Anthropic CLI`	Official Anthropic command-line interface for Claude models.
`Gemini CLI`	Google's command-line tool for interacting with Gemini models.

Comparison Table

Feature	gpt-cli	ollama	Claude Code	llm	shell_gpt	aichat
Model Support
OpenAI Models	✅	❌	❌	✅	✅	✅
Anthropic Models	✅	❌	✅	✅	❌	✅
Google Models	✅	❌	❌	✅	❌	✅
Grok Models	✅	❌	❌	❌	❌	❌
Groq Models	✅	❌	❌	✅	❌	✅
Together AI Models	✅	❌	❌	❌	❌	❌
Fireworks Models	✅	❌	❌	❌	❌	❌
MistralAI Models	✅	❌	❌	✅	❌	❌
Replicate Models	✅	❌	❌	✅	❌	❌
Local Models	✅	✅	❌	✅	❌	✅
Core Features
Interactive Chat	✅	✅	✅	✅	✅	✅
Single-prompt Mode	✅	✅	✅	✅	✅	✅
Pipeline Support	✅	✅	✅	✅	✅	✅
System Prompts	✅	✅	✅	✅	✅	✅
File Attachments	✅	✅	✅	✅	✅	✅
Model Switching	✅	✅	❌	✅	✅	✅
Technical Aspects
Language	TypeScript/Deno	Go	Python	Python	Python	Go
Local Execution	✅	✅	❌	✅	✅	✅
API Required	✅	❌	✅	✅	✅	✅
Open Source	✅	✅	❌	✅	✅	✅
Single binary operation	✅	✅	❌	❌	❌	✅
Integration
Vim Plugin	✅	❌	❌	❌	✅	❌
Shell Integration	✅	✅	✅	✅	✅	✅
Context Preservation	✅	✅	✅	✅	✅	✅
Unique Features
Model History Stack	✅	❌	❌	❌	❌	❌
In-chat Model Switching	✅	❌	❌	❌	❌	✅
Coding-Focused	❌	❌	✅	❌	❌	❌
Local Model Serving	❌	✅	❌	❌	❌	❌

Key Differences

gpt-cli vs. ollama

gpt-cli focuses on providing a unified interface for multiple cloud-based LLM providers, while ollama specializes in running models locally.
ollama includes its own model serving capability, while gpt-cli connects to existing API services (including ollama).
gpt-cli offers more flexibility in provider selection and model switching during conversations.

gpt-cli vs. Claude Code

gpt-cli is a general-purpose LLM client, while Claude Code is specifically designed for coding tasks.
Claude Code provides more coding-specific features and optimizations for developers.
gpt-cli supports a wider range of models beyond just Claude.

gpt-cli vs. llm/shell_gpt/aichat

gpt-cli is built with TypeScript/Deno, while most others use Python or Go.
gpt-cli offers unique features like model history stack and in-chat model switching.
Each tool has its own approach to command structure, configuration, and extension mechanisms.

When to Choose gpt-cli

When you need flexibility to switch between different LLM providers
When you want to use both cloud-based and local models
When you need Vim integration for code assistance
When you prefer a TypeScript/Deno-based solution
When you want to dynamically switch models during a conversation

How to Choose the Right Tool

For local model usage only: ollama might be more suitable
For dedicated coding tasks: Claude Code might be more specialized
For integration with Python ecosystems: llm or shell_gpt might be better
For maximum model flexibility: gpt-cli offers the widest range of options

Each tool has its strengths and ideal use cases. The best choice depends on your specific needs, workflow, and preferences.

Name		Name	Last commit message	Last commit date
Latest commit History 382 Commits
.github/workflows		.github/workflows
autoload		autoload
doc		doc
lib		lib
test		test
tools		tools
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
LICENSE		LICENSE
README.md		README.md
deno.json		deno.json
index.ts		index.ts
install.sh		install.sh
tsconfig.json		tsconfig.json

License

u1and0/gpt-cli

Folders and files

Latest commit

History

Repository files navigation

Quick start

Single mode (Not chat mode)

Use pipeline

Command Integration Examples

Installation

Download binary

Use runtime

Compile from source

Setup

API keys

OpenAI API (for GPT)

Anthropic API (for Claude)

Google API (for Gemini)

XAI API

Groq API

Together AI API

Fireworks API

MistralAI API

Replicate API

Setup Ollama (for Local running Models)

Usage

Options

PROMPT

Models

/ command

@ command

Test

Vim plugin (Optional)

LLM Terminal Clients Comparison

Common LLM Terminal Clients

Terminal-Based LLM Clients

Comparison Table

Key Differences

gpt-cli vs. ollama

gpt-cli vs. Claude Code

gpt-cli vs. llm/shell_gpt/aichat

When to Choose gpt-cli

How to Choose the Right Tool

About

Topics

Resources

License

Stars

Watchers

Forks

Releases 13

Packages 0

Languages

Packages