ReleaseOpenAIOpenAIpublished Sep 8, 2026seen 16h

openai/openai-agents-js v0.17.1

openai/openai-agents-js

Open original ↗

Captured source

source ↗
published Sep 8, 2026seen 16hcaptured 16hhttp 200method plain

v0.17.1

Repository: openai/openai-agents-js

Tag: v0.17.1

Published: 2026-09-08T10:04:10Z

Prerelease: no

Release notes:

What's Changed

  • feat: add server-wide MCP tool guardrails by @seratch in https://github.com/openai/openai-agents-js/pull/1754
  • feat: allow custom output guardrail blocked messages by @seratch in https://github.com/openai/openai-agents-js/pull/1745
  • feat(sandbox): allow labels on Docker sandbox containers by @seratch in https://github.com/openai/openai-agents-js/pull/1740
  • feat: support image results in web search tools by @seratch in https://github.com/openai/openai-agents-js/pull/1843
  • fix: enforce selected callable type alias contracts by @seratch in https://github.com/openai/openai-agents-js/pull/1749
  • fix: preserve serialized approval resume ownership by @seratch in https://github.com/openai/openai-agents-js/pull/1750
  • fix: recover failed resumed Session writes before continuing runs by @seratch in https://github.com/openai/openai-agents-js/pull/1757
  • fix: preserve Chat Completions tool turns during replay by @seratch in https://github.com/openai/openai-agents-js/pull/1766
  • fix: preserve approvals before handoffs and recover resumed writes by @seratch in https://github.com/openai/openai-agents-js/pull/1768
  • fix: prevent sandbox Git repository argument injection by @seratch in https://github.com/openai/openai-agents-js/pull/1840
  • fix: preserve tool-search attribution across agent handoffs by @seratch in https://github.com/openai/openai-agents-js/pull/1841
  • fix(core): honor EOF hunks for newline-terminated files by @seratch in https://github.com/openai/openai-agents-js/pull/1769
  • fix(core): wait for in-flight trace exports in forceFlush by @sylvesterkaczmarek in https://github.com/openai/openai-agents-js/pull/1813
  • fix(core): handle span processor rejections by @sylvesterkaczmarek in https://github.com/openai/openai-agents-js/pull/1796
  • fix(core): preserve CRLF line endings in applyDiff by @sylvesterkaczmarek in https://github.com/openai/openai-agents-js/pull/1806
  • fix(core): reject unsuccessful OpenAI Responses terminal states by @seratch in https://github.com/openai/openai-agents-js/pull/1720
  • fix(core): detach aggregated usage records by @seratch in https://github.com/openai/openai-agents-js/pull/1729
  • fix(core): preserve handoff callback errors by @sylvesterkaczmarek in https://github.com/openai/openai-agents-js/pull/1733
  • fix(core): isolate cached MCP tool definitions by @seratch in https://github.com/openai/openai-agents-js/pull/1727
  • fix(core): deduplicate MCP server lifecycle ownership by @seratch in https://github.com/openai/openai-agents-js/pull/1743
  • fix(core): strip created_by from RunItem replay by @seratch in https://github.com/openai/openai-agents-js/pull/1744
  • fix(openai): reject truncated empty Chat Completions by @seratch in https://github.com/openai/openai-agents-js/pull/1741
  • fix(openai): honor provider client options over defaults by @seratch in https://github.com/openai/openai-agents-js/pull/1728
  • fix(openai): preserve empty hosted MCP outputs by @Excelius-Wang in https://github.com/openai/openai-agents-js/pull/1747
  • fix(openai): restrict redacted response IDs to official endpoints by @seratch in https://github.com/openai/openai-agents-js/pull/1756
  • fix(openai): preserve response span data on early stream close by @seratch in https://github.com/openai/openai-agents-js/pull/1767
  • fix(openai): reset compaction state on clear by @sylvesterkaczmarek in https://github.com/openai/openai-agents-js/pull/1822
  • fix(openai): keep Conversations history limits local by @sylvesterkaczmarek in https://github.com/openai/openai-agents-js/pull/1791
  • fix(realtime): safely serialize guardrail feedback details by @sylvesterkaczmarek in https://github.com/openai/openai-agents-js/pull/1722
  • fix(extensions): decode PTY output incrementally by @pacocartones in https://github.com/openai/openai-agents-js/pull/1752
  • fix(openai): reset compaction state after pop by @sylvesterkaczmarek in https://github.com/openai/openai-agents-js/pull/1824
  • fix: preserve deferred tool identity across approval checkpoints by @seratch in https://github.com/openai/openai-agents-js/pull/1839
  • fix(sessions): preserve concurrent writes during automatic compaction by @seratch in https://github.com/openai/openai-agents-js/pull/1842
  • fix(sandbox): settle PTY output before cleanup by @seratch in https://github.com/openai/openai-agents-js/pull/1844
  • fix(sandbox): await PTY startup cleanup by @seratch in https://github.com/openai/openai-agents-js/pull/1845
  • fix(realtime): warn instead of silently dropping MCP items in resetHistory by @shoemoney in https://github.com/openai/openai-agents-js/pull/1748
  • refactor: share core Zod description lookup by @seratch in https://github.com/openai/openai-agents-js/pull/1834
  • refactor: isolate historical RunState compaction restoration by @seratch in https://github.com/openai/openai-agents-js/pull/1833
  • refactor: separate Responses item conversion from transport by @seratch in https://github.com/openai/openai-agents-js/pull/1836

Documentation & Other Changes

  • docs: document output guardrail replay safety by @seratch in https://github.com/openai/openai-agents-js/pull/1719
  • docs: clarify capability visibility and authorization by @seratch in https://github.com/openai/openai-agents-js/pull/1755
  • docs: explain Realtime WebRTC server-side controls by @seratch in https://github.com/openai/openai-agents-js/pull/1762
  • docs: correct typo in comments for agent tools description by @morozow in https://github.com/openai/openai-agents-js/pull/1774
  • docs: preserve canonical localized heading anchors by @seratch in https://github.com/openai/openai-agents-js/pull/1746
  • fix: route executable docs and contract changes to existing checks by @seratch in https://github.com/openai/openai-agents-js/pull/1828
  • fix: pin LF line endings so Windows clones keep fixture hashes by @Hashim1999164 in https://github.com/openai/openai-agents-js/pull/1721
  • fix: make test scripts work on Windows by @mikemikimike in https://github.com/openai/openai-agents-js/pull/1710
  • fix: enforce selected object property API contracts by @seratch in https://github.com/openai/openai-agents-js/pull/1847
  • perf: reduce Vitest global setup dependency loading by @seratch in https://github.com/openai/openai-agents-js/pull/1829
  • refactor: organize AI SDK message conversion and tests by @seratch in https://github.com/openai/openai-agents-js/pull/1830

*...

Excerpt shown — open the source for the full document.