Framework Integrations

Kibo is designed to support the ecosystem, not replace it. Here is how to configure agents for each supported backend, including Native Tool integration.

Agno (PhiData)

Best for general-purpose assistants and tool calling.

from agno.tools.yfinance import YFinanceTools

config = AgentConfig(
    agent="agno",
    model="gpt-4o-mini",
    config={
        "markdown": True,
        "show_tool_calls": False,
        "tools": [YFinanceTools()] # Pass native Agno tools
    }
)

LangChain

Uses ChatOpenAI and generic prompt templates. Supports langchain_core.tools and community tools.

from langchain_community.utilities import GoogleSerperAPIWrapper
from langchain_core.tools import Tool

search = GoogleSerperAPIWrapper()
tool = Tool(name="Search", func=search.run, description="Search Google")

config = AgentConfig(
    agent="langchain",
    model="gpt-4o-mini",
    config={
        "tools": [tool] # Pass native LangChain tools
    }
)

CrewAI

Creates a single-agent Crew (Agent + Task) for process-oriented workflows.

from crewai.tools import BaseTool

class MyTool(BaseTool):
    name: str = "MyTool"
    description: str = "Does something."
    def _run(self, arg): return "done"

config = AgentConfig(
    agent="crewai",
    model="gpt-4o-mini",
    config={
        "verbose": True,
        "tools": [MyTool()] # Pass native CrewAI tools
    }
)

PydanticAI

Uses PydanticAI's type-safe agent structure. Supports standard Python functions as tools.

def my_tool(x: int) -> int:
    """Returns double x."""
    return x * 2

config = AgentConfig(
    agent="pydantic-ai",
    model="gpt-4o-mini",
    config={
        "tools": [my_tool] # Pass python functions directly
    }
)

LangGraph

Deploys a pre-compiled state graph.

config = AgentConfig(
    agent="langgraph",
    model="gpt-4o-mini"
)