CrewAI

guard_tool wraps a CrewAI BaseTool so its arguments are scored by AxioRank before the tool executes; guard_tools does the same for a whole toolbelt. The tool's name, description, and args schema are untouched, so the guarded tool is a drop-in replacement anywhere the original was used.

Install

pip install "axiorank[crewai]"

Guard a crew's tools

from axiorank import AxioRank
from axiorank.integrations.crewai import guard_tools
from crewai import Agent

axio = AxioRank(api_key="axr_live_...")

researcher = Agent(
    role="Researcher",
    goal="Find and summarize sources",
    backstory="Methodical and careful.",
    tools=guard_tools([search_tool, scrape_tool, file_writer], axio, on_deny="return"),
)

Every _run now goes through your gateway first: allow runs the tool, deny blocks it, and a require_approval policy holds it for a human verdict before the crew continues.

Inspect what tools return

Pass inspect_results=True to also score the OUTPUT of untrusted-source tools (web fetches, file reads, email) for indirect prompt injection before the crew ingests it.

Next steps

• Framework integrations: the shared model and axio.trace().
• Python SDK: the underlying client.