OpenAI Agents SDK 0.21.0 ships provider-neutral testing utilities so agent workflows can be tested without any model call
Released 2026-08-15, v0.21.0 adds `agents.testing`, `agents.realtime.testing` and `agents.voice.testing` for deterministic Agent, Sandbox, Realtime and Voice workflow tests with no provider requests — scripted model utilities with frozen public API contracts, which is the piece most agent test suites have been hand-rolling with mocks. The release also moves provider compatibility to `openai>=3.0.0,<4` with HTTPX2-aware request/response/transport/exception handling, hardens RunState interruption snapshots, recursive agent-tool approvals, max-turn finalization, streaming cleanup and sensitive-error redaction, improves MCP lifecycle snapshot isolation with configurable retry backoff ceilings, and tightens view-image path grants in the Sandbox.
↳ Follow the thread