Direct Prompt (One-Shot CLI)

Run a one-shot AI task from the terminal with a single command — no YAML, no subcommand, just your prompt.

from praisonaiagents import Agent

agent = Agent(name="cli-agent", instructions="Process CLI prompts directly.")
agent.start("Answer: What is the capital of France?")

The user passes a bare prompt to praisonai; the dispatcher runs a one-shot agent without YAML or a subcommand.

Run an agent with a single bare command — the fastest way to get something done.

praisonai "summarise this folder"

How It Works

Quick Start

1. Install PraisonAI

   pip install praisonai
   
2. Set your API key

   export OPENAI_API_KEY="${OPENAI_API_KEY:?Set OPENAI_API_KEY in your shell}"
   
3. Run your first prompt

   praisonai "Hello"
   

   That’s it — the agent runs and returns a response immediately.

---

How Disambiguation Works

The CLI inspects the bare positional argument and routes it using these rules (evaluated in order):

Positional value	Routed as
An existing file path on disk	Agent file
Ends in .yaml or .yml (any case)	Agent file
A known subcommand (run, chat, memory, …)	Subcommand
Anything else (e.g. "summarise this folder")	One-shot direct prompt

When to use praisonai run instead: the bare form is a shortcut for simple one-off prompts. Use praisonai run when you need flags such as --output stream-json, --model, --continue, --session, --no-rules, --allow, etc. See praisonai run for the full flag reference.

---

Examples

praisonai "What is the capital of France?"
praisonai "Summarize the README.md file"
praisonai "Write a Python function to reverse a string"
praisonai "List the top 5 AI trends in 2025"

---

Best Practices

Always quote multi-word prompts

Wrap your prompt in double quotes to prevent the shell from splitting it into multiple arguments:

praisonai "summarise this folder"   # correct
praisonai  summarise this folder    # wrong — shell sees separate args

Avoid name collisions with local files

If a file named summarise exists in your current directory, the CLI treats the positional as a file path. Quote the prompt and ensure it doesn’t match a real file name, or use praisonai run "summarise" which forces prompt mode.

Combining stdin with a positional prompt

You can pipe content into the CLI and provide a prompt at the same time:

cat report.txt | praisonai "Summarise the following text"

The piped stdin is appended to the agent’s context alongside the prompt.

Using with different models

The bare positional form works with any model flag:

praisonai run "Explain quantum computing" --model gpt-4o

The --model flag requires praisonai run — the bare shortcut form does not accept flags.

---

Related

praisonai run

Full flag reference for the run subcommand, including output modes, session continuity, and permission controls.

Quick Start

End-to-end guide for installing PraisonAI and running your first agent.