Skip to main content
Agent Recipes are ready-to-use AI agent blueprints. Think of them like cooking recipes - follow the steps and get reliable results every time.

Why Use Recipes?

No Coding Required

Just run them - everything is pre-configured

Battle-Tested

Community-proven patterns that work

Customizable

Adjust variables to fit your needs

Quick Start

1

Install PraisonAI

pip install praisonai
2

Set Your API Key

export OPENAI_API_KEY=your_key_here
3

Run a Recipe

praisonai recipe run research-agent --var topic="AI trends 2024"

How Recipes Work

StepWhat Happens
1. You provide inputTell the recipe what to work on
2. Agents are createdThe blueprint builds specialized workers
3. Tasks executeEach agent completes their job
4. Tools helpAgents use search, files, etc.
5. You get resultsCombined output from all agents

Research Agent

Searches the web and creates summaries on any topic
praisonai recipe run research-agent --var topic="Your topic"

Video Editor

AI-powered video editing and optimization
praisonai recipe run ai-video-editor --var input=video.mp4

Code Reviewer

Reviews code and suggests improvements
praisonai recipe run code-reviewer --var repo=./myproject

Content Writer

Creates articles, blog posts, and marketing copy
praisonai recipe run content-writer --var topic="Your topic"

Recipe Commands

What you want to doCommand
See all recipespraisonai recipe list
Get recipe detailspraisonai recipe info <name>
Run a recipepraisonai recipe run <name>
Run with optionspraisonai recipe run <name> --var key=value
Create your ownpraisonai recipe init my-recipe

Anatomy of a Recipe

Every recipe is just a folder with a configuration file:
my-recipe/
├── agents.yaml    ← The blueprint
├── tools.py       ← Optional: custom tools
└── README.md      ← Optional: documentation

The agents.yaml File

This is where the magic happens. Here’s a simple example:
framework: praisonai
topic: "{{task}}"

roles:
  researcher:
    role: Research Specialist
    goal: Find accurate information
    tools:
      - internet_search
    tasks:
      research:
        description: "Research {{task}} and provide key findings"
        expected_output: "Summary with sources"
The {{task}} is a variable - you fill it in when running the recipe!

Customizing Recipes

Option 1: Change Variables

Most recipes have built-in options you can adjust:
# See what options are available
praisonai recipe info research-agent

# Run with your settings
praisonai recipe run research-agent \
  --var topic="Renewable energy" \
  --var depth="comprehensive"

Option 2: Edit the Blueprint

For bigger changes, modify the agents.yaml file:
  1. Find or create a recipe folder
  2. Edit the agents.yaml file
  3. Run your modified recipe
# agents.yaml - add another agent!
roles:
  researcher:
    role: Researcher
    goal: Find information
    tasks:
      research:
        description: "Research {{topic}}"
  
  writer:  # ← Add this!
    role: Writer
    goal: Write engaging content
    tasks:
      write:
        description: "Write about the research findings"

Creating Your Own Recipe

1

Initialize

praisonai recipe init my-awesome-recipe
cd my-awesome-recipe
2

Edit agents.yaml

Define your agents and what they should do:
framework: praisonai
topic: "{{task}}"

roles:
  helper:
    role: Helpful Assistant
    goal: Complete the user's task
    tasks:
      main:
        description: "{{task}}"
        expected_output: "Task completed successfully"
3

Test It

praisonai recipe run . --var task="Say hello world"

Recipe Variables

Variables let you customize recipes without editing files:
In the YAML fileHow to use it
{{topic}}--var topic="Your topic"
{{input}}--var input="file.txt"
{{output}}--var output="result.txt"
Variable names are flexible - recipe authors can name them anything!

Adding Tools to Recipes

Tools give your agents superpowers:
roles:
  agent:
    role: My Agent
    tools:
      - internet_search    # Search the web
      - file_read_tool     # Read files
      - shell_tool         # Run commands

Available Tools

ToolWhat it does
internet_searchSearch the web for information
file_read_toolRead files from disk
file_write_toolWrite files to disk
shell_toolRun terminal commands
Run praisonai tools list to see all available tools!

Running Recipes from GitHub

Share and use recipes from anywhere:
# Run directly from GitHub
praisonai recipe run github:MervinPraison/Agent-Recipes/research-agent

# From a specific branch
praisonai recipe run github:MervinPraison/Agent-Recipes/research-agent@main

Troubleshooting

Make sure you’re in the right directory or use the full path:
praisonai recipe run ./my-recipe
Check if the tool is available:
praisonai tools list
Make sure you’re using the correct variable name with --var:
praisonai recipe run my-recipe --var topic="My topic"
Enable verbose mode to see what’s happening:
praisonai recipe run my-recipe --verbose

Best Practices

Browse the recipe library first - someone may have already built what you need!
Start with one agent, then add more as needed. Simpler is usually better.
Tell your agents exactly what you want in the task description.
Test your recipe after each change to catch issues early.