Stop switching contexts. Many AI CLI tools are heavy, Node.js-based, and trap you in their interface. They might go back and forth just to run a simple ls, cutting off real access to your command line.
pls is different: lightweight, fast, and built for everyday CLI tasks, keeping you fully in the command line while letting you seamlessly switch between AI and shell commands.
- You are working on a file, and use
plsto help process it:
- You
catthe result, are happy with it, and move on something else - You logout ... and login back
- Then you
cdto another working folder where you want to continue the task, - By simple type
pls, the AI picks up the task, and through interative chats you provide new input:
- The AI generated shell commands for you to run,
- And you decided to further tweek the commands,
- Finally, the command runs successfully.
- Use
cat,tail, or any command pipe to feed data intopls. - You can also feed "knowhow" like
my_app --help | pls learn it - Or generate data for your app
pls generate some markdown demo | python my_markdown_render.py - Give orders in chat mode, refine your request as many times until satisfied.
- Quit chat at any time to run any command you want, and pick it up again anywhere, anytime, by typing
pls. - Use AI for its flexibility to handle anything; for simple commands, run them yourself for efficiency.
- For short context, let AI do it directly; for large datasets, ask AI to generate commands or scripts for you to run.
- If not satisfied with anything, simply provide your feedback.
- You can apply the workflow to tasks like system administration, config file tweaking, log analysis, and script writing and testing.
- It will make your command-line work easier and more fun, while still keeping the traditional command-line style and feel.
- Generate chat messages and shell commands in a single request, it never “thinks twice.”
- Switch effortlessly between interactive mode and command mode with shared memory.
- Keep a short-term history, allowing you to pick up anytime without session management.
- Run as a single Bash script, with no extra tools or MCP needed, consumes very few tokens.
- (Experimental) Configure via chat:
update yourself,edit config,delete chat history.
- AI chat and shell command generation
- Inline command regeneration and editing
- Supports pipes and command chaining, clean output in piped mode
- Works on Linux and macOS
- Compatible with OpenAI API and Gemini API
pls v0.56
Usage: pls [messages...] # Chat with an input
> what is llm # Continue chat, q or empty input to quit
Examples: pls # Start without input
pls count files # ls -1 | wc -l # shell cmd wait for run
> include subdirs # find . -type f | wc -l # shell cmd update
Pipe and Chain:
echo how to cook rice | pls # Use input from pipe
echo rice | pls how to cook # Args + pipe (task from args, data from pipe)
pls name a dish | pls -p how to cook # Chain commands and show piped input with -p
Settings: pls -h # Show this help
pls edit config # config pls and AI model via chat
- install
jqif not installed already - run the follow script to install or update
curl -sSL https://raw.githubusercontent.com/cjccjj/pls/main/install.sh | bash- Export
OPENAI_API_KEYorGEMINI_API_KEYin your shell config - Optional: install
glowfor prettier markdown output in chat
- To config and choose AI profile say
edit configto pls ornano ~/.pls/pls.conf
- Say
update yourselfor use install script
- run
sudo rm /usr/local/bin/pls # remove the bin
rm -rf ~/.pls # remove config file and chat history