antarraksha-openai-agents

Antarraksha AI Agent Enforcement SDK for the OpenAI Agents SDK.

Installation

pip install antarraksha-openai-agents

Quick Start

Attach Antarraksha as input and output guardrails on your Agent. Registration happens automatically on first use — no login, no API key, no signup required.

from agents import Agent
from antarraksha_openai_agents import (
    AntarrakshaInputGuardrail,
    AntarrakshaOutputGuardrail,
)

input_guard = AntarrakshaInputGuardrail(agent_id="my-openai-agent")
output_guard = AntarrakshaOutputGuardrail(agent_id="my-openai-agent")

agent = Agent(
    name="my-openai-agent",
    instructions="You are a helpful assistant.",
    input_guardrails=[input_guard],
    output_guardrails=[output_guard],
)

Run your agent normally — every input and output is now enforced against Antarraksha policy.

Parameters

Enforcement Behavior

• Input guardrail fires before the LLM receives the user prompt. Antarraksha can DENY (raises PermissionError) or ALLOW.
• Output guardrail fires after the LLM produces output. Antarraksha can DENY (raises PermissionError) or ALLOW.
• Every call is logged in your Antarraksha registry and visible at https://antarraksha.ai/registry.

Corporate Network Note

If you're behind a corporate TLS-inspection proxy and see SSLCertVerificationError, install:

pip install pip-system-certs

This makes Python trust your Windows / macOS system certificate store.