OpenAI Agents SDK
AI governance for the OpenAI Agents SDK — input and output guardrails with audit trails for every agent interaction.
OpenAI Agents SDK Integration
Input and output guardrails for the official OpenAI Agents framework. Checks every agent interaction against your Palveron governance policies. Blocked requests raise native GuardrailTripwireTriggered exceptions.
Installation
pip install palveron-openai-agentsQuickstart
from agents import Agent
from palveron_openai_agents import palveron_input_guardrail, palveron_output_guardrail
agent = Agent(
name="assistant",
instructions="You are a helpful assistant.",
input_guardrails=[palveron_input_guardrail(api_key="pv_live_xxx")],
output_guardrails=[palveron_output_guardrail(api_key="pv_live_xxx")],
)Inputs with PII are blocked before the agent processes them. Outputs with secrets are blocked before the user sees them.
Guardrails
| Guardrail | When | On BLOCKED |
|---|---|---|
palveron_input_guardrail | Before agent processing | Raises GuardrailTripwireTriggered |
palveron_output_guardrail | Before output to user | Raises GuardrailTripwireTriggered |
Configuration
guardrail = palveron_input_guardrail(
api_key="pv_live_xxx",
base_url="https://gateway.internal.company.com:8080",
fail_open=False,
metadata={"team": "support"},
)Source Code
Open source (MIT): github.com/palveron/adapter-openai-agents.
Next Steps
- LangChain integration for pipeline governance
- CrewAI integration for multi-agent governance
- Create your own policies
CrewAI
AI governance for CrewAI — task guardrails, step callbacks, and audit trails for every crew run.
Google ADK
AI governance for the Google Agent Development Kit — automatic policy checks for every tool call, model invocation, and A2A message, with multimodal-aware extraction and MCP context forwarding.