Low Code Custom Tools

Step-by-step guide for creating and implementing custom tools in PraisonAI, including examples and configuration instructions

from praisonaiagents import Agent, tool

@tool
def weather(city: str) -> str:
    """Return a short weather summary for a city."""
    return f"Sunny in {city}"

agent = Agent(name="Assistant", tools=[weather])
agent.start("What is the weather in London?")

The user asks in plain language; custom tools extend what the agent can do beyond built-ins.

Create Custom Tools

Quick Start

1. Install

   pip install praisonaiagents
   
2. Create a custom tool

   from praisonaiagents import Agent
   
   def get_weather(city: str) -> str:
       return f'Weather in {city}: Sunny, 22 degrees'
   
   agent = Agent(
       name="WeatherAgent",
       instructions="Provide weather information using the weather tool.",
       tools=[get_weather],
   )
   
   agent.start("What is the weather in London?")
   

Step 1: Install the praisonai Package

First, you need to install the praisonai package. Open your terminal and run the following command:

pip install praisonai

Step 2: Create the InternetSearchTool

Next, create a file named tools.py and add the following code to define the InternetSearchTool:

from duckduckgo_search import DDGS
from praisonai_tools import BaseTool

class InternetSearchTool(BaseTool):
    name: str = "Internet Search Tool"
    description: str = "Search Internet for relevant information based on a query or latest news"

    def _run(self, query: str):
        ddgs = DDGS()
        results = ddgs.text(keywords=query, region='wt-wt', safesearch='moderate', max_results=5)
        return results

Step 3: Define the Agent Configuration

Create a file named agents.yaml and add the following content to configure the agent:

framework: crewai
topic: research about the causes of lung disease
agents:  # Canonical: use 'agents' instead of 'roles'
  research_analyst:
    instructions:  # Canonical: use 'instructions' instead of 'backstory' Experienced in analyzing scientific data related to respiratory health.
    goal: Analyze data on lung diseases
    role: Research Analyst
    tasks:
      data_analysis:
        description: Gather and analyze data on the causes and risk factors of lung diseases.
        expected_output: Report detailing key findings on lung disease causes.
    tools:
    - InternetSearchTool

Step 4: Run the PraisonAI Tool

To run the PraisonAI tool, simply type the following command in your terminal:

praisonai

If you want to run the AG2 framework (Formerly AutoGen), use:

praisonai --framework autogen

Prerequisites

Ensure you have the duckduckgo_search package installed. If not, you can install it using:

pip install duckduckgo_search

That’s it! You should now have the PraisonAI tool installed and configured.

Other information

TL;DR to Create a Custom Tool

pip install praisonai duckduckgo-search
export OPENAI_API_KEY="Enter your API key"
praisonai --init research about the latest AI News and prepare a detailed report

• Add - InternetSearchTool in the agents.yaml file in the tools section.
• Create a file called tools.py and add this code tools.py

praisonai

Pre-requisite to Create a Custom Tool

agents.yaml file should be present in the current directory.

If it doesn’t exist, create it by running the command praisonai --init research about the latest AI News and prepare a detailed report.

Step 1 to Create a Custom Tool

Create a file called tools.py in the same directory as the agents.yaml file.

# example tools.py
from duckduckgo_search import DDGS
from praisonai_tools import BaseTool

class InternetSearchTool(BaseTool):
    name: str = "InternetSearchTool"
    description: str = "Search Internet for relevant information based on a query or latest news"

    def _run(self, query: str):
        ddgs = DDGS()
        results = ddgs.text(keywords=query, region='wt-wt', safesearch='moderate', max_results=5)
        return results

Step 2 to Create a Custom Tool

Add the tool to the agents.yaml file as show below under the tools section - InternetSearchTool.

framework: crewai
topic: research about the latest AI News and prepare a detailed report
agents:  # Canonical: use 'agents' instead of 'roles'
  research_analyst:
    instructions:  # Canonical: use 'instructions' instead of 'backstory' Experienced in gathering and analyzing data related to AI news trends.
    goal: Analyze AI News trends
    role: Research Analyst
    tasks:
      gather_data:
        description: Conduct in-depth research on the latest AI News trends from reputable
          sources.
        expected_output: Comprehensive report on current AI News trends.
    tools:
    - InternetSearchTool

---

Python Code Custom Tools

For praisonaiagents package users, you can create custom tools using the @tool decorator or BaseTool class.

Using @tool Decorator

from praisonaiagents import Agent, tool

@tool
def search(query: str) -> str:
    """Search the web for information."""
    return f"Results for: {query}"

@tool
def add(a: float, b: float) -> float:
    """Add two numbers (safe — no eval)."""
    return a + b

agent = Agent(
    instructions="You are a helpful assistant",
    tools=[search, add]
)
agent.start("Search for AI news and add 15 and 4")

Using BaseTool Class

from praisonaiagents import Agent, BaseTool

class WeatherTool(BaseTool):
    name = "weather"
    description = "Get current weather for a location"
    
    def run(self, location: str) -> str:
        return f"Weather in {location}: 72°F, Sunny"

agent = Agent(
    instructions="You are a weather assistant",
    tools=[WeatherTool()]
)
agent.start("What's the weather in Paris?")

Expressive Parameter Types

Create tools with rich type annotations for better schema generation:

from typing import Optional, Literal
from praisonaiagents import Agent, tool

@tool
def configure_service(
    name: str,
    port: Optional[int] = None,
    mode: Literal["production", "development"] = "development"
) -> str:
    """Configure a service with optional port and mode."""
    config = f"Service '{name}' configured in {mode} mode"
    if port:
        config += f" on port {port}"
    return config

agent = Agent(
    instructions="You configure services",
    tools=[configure_service]
)
agent.start("Configure webapp in production mode with port 8080")

See the Tool Parameter Types page for complete guide.

Creating a Pip-Installable Tool Package

Create tools that auto-register when installed:

pyproject.toml

[project]
name = "my-praisonai-tools"
version = "1.0.0"
dependencies = ["praisonaiagents"]

[project.entry-points."praisonaiagents.tools"]
my_tool = "my_package:MyTool"

my_package/init.py

from praisonaiagents import BaseTool

class MyTool(BaseTool):
    name = "my_tool"
    description = "My custom tool"
    
    def run(self, param: str) -> str:
        return f"Result: {param}"

Usage

After pip install, tools are auto-discovered:

from praisonaiagents import Agent

agent = Agent(tools=["my_tool"])  # Works automatically!

---

What Happens During Validation

The @tool decorator and BaseTool instances now validate schemas automatically to ensure OpenAI compatibility.

Automatic validation when using @tool:

@tool
def search(query: str) -> str:
    """Search the web for information."""
    return f"Results for: {query}"
# Validation runs here - check logs for warnings

Important notes:

• The @tool decorator validates at creation time and logs warnings (doesn’t raise)
• parameters (if set manually) must include properties for OpenAI compatibility
• Validation failures are logged - scan your logs for Tool validation warning for ...

Example that will raise ToolValidationError at agent creation:

# This will raise ToolValidationError at agent creation:
class BadTool(BaseTool):
    name = "bad"
    description = "..."
    parameters = {"type": "object"}  # missing 'properties'
    def run(self, q: str) -> str: return q

For comprehensive validation details and error fixes, see Tool Schema Validation.

Tool Name Collisions: When developing custom tools, avoid naming conflicts with existing tools. If you have multiple environments with overlapping tool names, use the Allowed Tools feature to whitelist specific tools.

Best Practices

Use simple function signatures

Define tools as plain Python functions with clear type hints - def my_tool(query: str) -> str.

Write descriptive docstrings

The agent reads tool docstrings to decide when to call them - be specific about what the tool does.

Always return strings

Return a string from tool functions - even for errors - so the agent can process the result.

Test tools independently

Call your tool function directly with test inputs before adding it to an agent.

Related

Custom Tools

Build your own agent tools

Tools Overview

Browse PraisonAI tool documentation