Directory / OpenAI Agents SDK

OpenAI Agents SDK

Agent framework · OpenAI · 3/7 verified · Last checked 2026-10-16

needs_approval on tools pauses a run and returns interruptions you approve or reject. Built-in tracing records tool calls and handoffs. result.cancel() stops a streaming run.

Overall: C (2.67 avg, 3/7 controls verified; capped from A because A needs 5 verified)

Lock this down first

Mark every write tool needs_approval=True and keep sensitive-data tracing off in production.

Scorecard

Criterion	Rating	Notes
Permission granularity	Unverified	Not confirmed in the docs we checked.
Human approval gates	3/3	Tools set needs_approval (true or a function). Runs pause with interruptions you approve or reject.
Sandboxing	Unverified	Not confirmed in the docs we checked.
Secrets handling	Unverified	Not confirmed in the docs we checked.
Audit logs	3/3	Tracing covers model calls, tool calls, handoffs and guardrails.
Kill switch / interrupt	2/3	result.cancel() stops a streaming run. Guardrail tripwires raise an exception, but in parallel mode tools may already have run.
Network egress control	Unverified	Not confirmed in the docs we checked.

Unverified: Permission granularity, Sandboxing, Secrets handling, Network egress control. That means we could not confirm it in public docs, not that it is missing.

Sources

Vendor? Request a correction. We review corrections within 7 days.