Langfuse Integration

Note: Control Zero handles policy enforcement. Langfuse is a complementary observability tool. For tracing and evaluation, these tools work alongside Control Zero.

Use Control Zero for policy enforcement alongside Langfuse for tracing.

Overview

Control Zero and Langfuse solve different problems and work well together:

• Control Zero: Enforces governance policies (what agents are allowed to do)
• Langfuse: Provides observability and tracing (what agents actually did)

Together, they give you both prevention (block unauthorized actions) and visibility (trace and debug all actions).

Installation

pip install controlzero langfuse openai

Combined Setup

import controlzero
from langfuse import Langfuse
from langfuse.openai import openai  # Langfuse-wrapped OpenAI client

# -- Control Zero: policy enforcement --
cz = controlzero.ControlZero()
cz.initialize()

# -- Langfuse: observability --
langfuse = Langfuse(
    public_key="pk-...",
    secret_key="sk-...",
    host="https://cloud.langfuse.com",
)

Usage Pattern

Enforce policies with Control Zero, then let Langfuse trace the execution:

from langfuse.decorators import observe

@observe()
def generate_response(prompt: str, agent_id: str) -> str:
    """Generate a response with both enforcement and tracing."""

    # Step 1: Control Zero enforces the policy
    cz.enforce(
        action="llm.generate",
        resource="model/gpt-4",
        context={"agent_id": agent_id},
    )

    # Step 2: Langfuse traces the call (via its OpenAI wrapper)
    response = openai.chat.completions.create(
        model="gpt-4",
        messages=[{"role": "user", "content": prompt}],
    )

    return response.choices[0].message.content


@observe()
def handle_tool_call(tool_name: str, arguments: dict, agent_id: str) -> str:
    """Execute a tool call with enforcement and tracing."""

    # Enforce before execution
    cz.enforce(
        action="tool.call",
        resource=f"tool/{tool_name}",
        context={"agent_id": agent_id},
    )

    # Tool execution is automatically traced by Langfuse
    result = execute_tool(tool_name, arguments)
    return result

What Each Tool Provides

Capability	Control Zero	Langfuse
Block unauthorized actions	Yes	No
Allow/deny policies	Yes	No
Encrypted policy bundles	Yes	No
LLM call tracing	No (logs decisions only)	Yes
Cost tracking	No	Yes
Prompt management	No	Yes
Latency monitoring	No	Yes
Evaluation scores	No	Yes

Next Steps

• See the Python SDK for the full Control Zero API reference.
• Learn about Policies for defining governance rules.