LangChain, CrewAI and LlamaIndex

The adapters ship with the Python package (pip install zn-gate) under zn_gate.integrations. They use local rules by default and the cloud model when ZN_API_KEY is set.

LangChain and LangGraph

A callback handler checks prompts and tool inputs as the agent runs:

from zn_gate.integrations import ZnGuardCallbackHandler

handler = ZnGuardCallbackHandler(raise_on_injection=True)
agent = initialize_agent(tools, llm, callbacks=[handler])

CrewAI

Wrap a tool function so blocked calls return a fallback instead of executing:

from zn_gate.integrations import guarded_tool

@guarded_tool(on_block="return", fallback="BLOCKED_BY_GUARD", mask_secrets=True)
def run_search(query: str):
    return search_api(query)

LlamaIndex

Check a query before it reaches the index:

from zn_gate.integrations import ZnLlamaGuard

guard = ZnLlamaGuard(raise_on_violation=True)
safe_query = guard.on_query_start(user_query)

Not using a framework? See the SDK.