Claude AI Toolkit

Unmaintained. Use the anthropic SDK for Python or Claude Code for CLI.

I've always preferred pair programming with LLMs from the terminal over copy-pasting from browser chats. Web UIs hit rate limits, lose context across tabs, and break flow when I'm moving code back and forth. Anthropic dropped the Claude 3 family on March 4, 2024. I shipped this the same day. The official Python SDK was fresh, Claude Code didn't exist yet, and I wanted chat/text/vision modes, streaming, system prompts, and the full parameter surface behind one CLI. So I built this for myself.

Python wrapper and CLI for Anthropic's Claude models, with vision.

Install

git clone https://github.com/ramonclaudio/claude-ai-toolkit.git
cd claude-ai-toolkit
pip install -r requirements.txt

Configuration

Get an API key at https://console.anthropic.com/. Set via env var, .env, or pass directly:

export CLAUDE_API_KEY=your_api_key

CLI

# Chat
python cli.py --chat

# Text
python cli.py --text --prompt "Write a haiku about robots."

# Vision
python cli.py --vision --prompt "Describe this image." --image "https://example.com/image.jpg"

Type exit or quit to leave chat.

Python

from claude import Chat, Text, Vision

Chat().run()
Text().run(prompt="Write a haiku about robots.")
Vision().run(prompt="Describe this image.", image="https://example.com/image.jpg")

Options

Description	CLI	Python
Chat mode	--chat	Chat()
Text mode	--text	Text()
Vision mode	--vision	Vision()
Prompt	--prompt	prompt=
Image path or URL	--image	image=
API key	--api_key	api_key=
Model	--model	model=
Streaming	--stream	stream=True
System prompt	--system_prompt	system_prompt=
Max tokens	--max_tokens	max_tokens=
Temperature	--temperature	temperature=
Top-p	--top_p	top_p=
Top-k	--top_k	top_k=
Stop sequences	--stop_sequences	stop_sequences=

Models

Model	Max tokens
claude-3-5-sonnet-20241022	4096
claude-3-opus-20240229	4096
claude-3-haiku-20240307	4096

Vision

Supports JPEG, PNG, GIF, WEBP. If the image's long edge is over 1568 pixels or the image exceeds ~1600 tokens, Anthropic will auto-resize it (preserving aspect ratio). Very small images under 200 pixels on any edge may hurt performance.

License

MIT