OpenAI Agents SDK
Orchestrate GPT-powered agents, tools, handoffs, guardrails and sessions with built-in tracing.
Official product pageAvery Rulebook for GPT agents
Keep the agent loop lightweight. Keep enterprise policy authoritative.
The OpenAI Agents SDK provides agents, tools, handoffs, guardrails, sessions, human review and tracing. Rulebook adds a model-independent enterprise policy artifact that many GPT agents and non-OpenAI workflows can share.
The operating reality
SDK guardrails validate agent inputs and outputs, while enterprise policy may need to resolve product eligibility, delegated authority, contractual obligations and jurisdiction before a tool call or recommendation can proceed.
Keepthe product you chose
Addexecutable enterprise policy
Provewhy every action was allowed
The runtime compliance gate
Check the answer or action before it creates impact.
Rulebook evaluates live context against the confirmed policy and regulatory rules that apply, then returns a decision the agent must follow.
01Context
Actor, purpose, data, jurisdiction and proposed action
02Gate
Allow, deny, obligations or named human review
03Audit
Evidence, rulebook version and signed decision trace
Architecture boundary
Avery Rulebook
Centralize confirmed business-policy logic and return structured allow, deny, obligation or human-review decisions.
The Agents SDK owns the agent runtime and execution trace. Rulebook owns enterprise-policy compilation, publication and decision evidence.
Why the gap persists
Policies exist. Runtime compliance gates do not.
01
Input and output validation does not answer every business-authorization question.
02
Policy logic can drift when copied into many agents and tool handlers.
03
Agent traces need the governing rule version to explain why an action was permitted.
Rulebook in the workflow
Run the agent
The SDK manages the turn, context, handoffs and proposed tool invocation.
Call policy
A function tool, MCP tool or approval handler requests a Rulebook decision.
Join the evidence
The application links the Rulebook receipt with the OpenAI trace and resulting business event.
Potential outcomes
Make compliance a gate in the work, not a report about the work.
- One confirmed policy across many GPT agents
- Clear separation between SDK guardrails and enterprise rules
- End-to-end evidence from agent trace to business authority
Questions leaders ask
Why not implement every rule as an Agents SDK guardrail?
SDK guardrails are valuable for agent validation. Rulebook is designed for policy owned by enterprise experts, shared across runtimes, versioned independently and backed by source evidence.
Can Rulebook participate in human-in-the-loop flows?
Yes. Rulebook can specify when review is required and which authority is needed, while the Agents SDK pauses and resumes the run.
Executive briefing
Add an auditable compliance gate to OpenAI Agents SDK integration.
Bring one consequential agent workflow. We will map the governing policies and regulations, runtime gate, human authority and auditable evidence path with your team.
Request a briefing