CrewAI - Keenable

Agent framework Give CrewAI agents Keenable web search with a small tool that wraps the official Python SDK. You can try it without an API key. Set KEENABLE_API_KEY to use your organization’s limits. CrewAI’s own KeenableSearchTool is in review upstream (PR #7570). Until it ships, this page wraps the SDK.

Install

pip install keenable 'crewai[tools]'

Add search to a crew

Define a tool that calls the SDK, then give it to an agent:

from crewai import Agent, Task, Crew
from crewai.tools import tool
from keenable import Keenable

client = Keenable()  # keyless by default; reads KEENABLE_API_KEY when set


@tool("Keenable Web Search")
def keenable_search(query: str) -> str:
    """Search the web and return cited results for the agent."""
    return client.search(query, max_results=5).to_context()


researcher = Agent(
    role="Researcher",
    goal="Answer with live web sources",
    backstory="You search before you claim a fact.",
    tools=[keenable_search],
)

task = Task(
    description="What changed in the latest Python release? Cite sources.",
    expected_output="A short answer with numbered source URLs.",
    agent=researcher,
)

Crew(agents=[researcher], tasks=[task]).kickoff()

to_context() returns a numbered block of [n] Title (url) entries that the model can cite. To read a single page, add a tool that calls page = client.fetch(url) and returns page.content.

Set KEENABLE_API_KEY to remove the hourly request cap. If you don’t set a key, calls run on the shared public tier at lower rate limits.