Pre-tool guardrail · PyPI 0.1.0 · Alpha · Apache-2.0

openai-agents-acgs for OpenAI Agents SDK.

Applies tool input guardrails and needs_approval checks to local function tools in the OpenAI Agents SDK.

PyPI 0.1.0 GitHub repository Inspect src/openai_agents_acgs/guard.py#L36

Package Specification · Inspected HEAD f2cd08c

Framework maintainer
OpenAI · openai.github.io
Upstream repository
openai/openai-agents-python
PyPI release
openai-agents-acgs==0.1.0
Python requirement
>=3.10
Framework constraint
openai-agents>=0.22.1,<1
Governance engine pin
acgs-lite>=3.0.1,<4
Integration mechanic
Pre-tool veto / intervention hook
Hook entry point
ACGSToolGuard.govern_tool / govern_agent

01 · Local Setup & Quickstart

Attach governance to one tool-calling agent.

Install the pinned release from PyPI alongside acgs-lite>=3.0.1,<4. Rule evaluation uses deterministic keyword and regex matching locally in Python—no external model or network call is made by the governance check.

Caveat: Provider-hosted tools (such as hosted web search or file search) execute outside local Python hooks and are not governed by this adapter.

Shell · PyPI installation

pip install "openai-agents-acgs==0.1.0"

Python · OpenAI Agents SDK integration

src/openai_agents_acgs/guard.py#L36
from agents import Agent, function_tool
from openai_agents_acgs import ACGSToolGuard


@function_tool
def send_email(to: str, body: str) -> str:
    """Send an email."""
    return f"sent to {to}"


guard = ACGSToolGuard()
agent = guard.govern_agent(Agent(name="Assistant", tools=[send_email]))

02 · Verdict & Review Mapping

How ACGS-Lite verdicts map to OpenAI Agents SDK.

When a tool call is proposed, the adapter formats the tool name and arguments via format_tool_call and evaluates them against the configured Constitution. Blocking rules always take precedence over review rules.

Verdict.ALLOW

Returns ToolGuardrailFunctionOutput.allow() so the SDK proceeds with FunctionTool execution.

Verdict.DENY

Returns ToolGuardrailFunctionOutput.reject_content(...) (or triggers ToolInputGuardrailTripwireTriggered when on_deny='raise').

Verdict.REVIEW

Triggers needs_approval to pause the run as a ToolApprovalItem interruption on RunState.

03 · Public API Surface

Exported symbols and source entry points.

Each adapter keeps its public surface minimal and re-exports the shared GovernanceDecision, Verdict, and ToolCallFormatter types alongside the framework-specific hook or middleware class.

__all__ exports · openai-agents-acgs

src/openai_agents_acgs/guard.py#L36 ↗
ACGSToolGuardApprovalPolicyGUARDRAIL_NAMEUngovernedToolsErrorGovernanceDecisionToolCallFormatterVerdictformat_tool_call

Primary hook surface: ACGSToolGuard.govern_tool / govern_agent. Custom ToolCallFormatter functions may be supplied if your tool arguments require custom canonical serialization before rule matching.

04 · Scope & Bypass Boundaries

What this adapter does and does not cover.

Verify these framework-specific boundaries when integrating openai-agents-acgs into an evaluation workflow.

All 13 adaptersReceipts, grants & EffectStore
  1. 01

    Because the OpenAI Agents SDK has no run-wide tool interception hook, you must wrap tools with govern_agent(agent) or govern_tool(tool).

  2. 02

    Provider-hosted tools (WebSearchTool, FileSearchTool, ComputerTool, CodeInterpreterTool, HostedMCPTool) cannot attach FunctionTool guardrails and raise UngovernedToolsError when passed to govern_agent.

  3. 03

    In the current 0.1.x adapter series, rule evaluation runs through acgs-lite's ToolCallGovernor path. If your workflow requires SignedReceipt, ExecutionGrant, or SQLite EffectStore reservations, wire those contracts explicitly at the tool executor boundary.

Other pre-tool guardrail packages

claude-agent-acgs0.1.0

Claude Agent SDK · Anthropic

pydantic-ai-acgs0.1.0

Pydantic AI · Pydantic

adk-acgs0.1.1

Google ADK · Google

crewai-acgs0.1.0

CrewAI · CrewAI

Put one real workflow under your control.

Choose one tool. Define what it may do. Inspect what happens when a request crosses the boundary.

Request evaluation access Explore developer tools Compare all products Review Claims & Evidence