Issues

• Nested handoff history: speaker names in summary records break Chat Completions agents when a handoff_history_mapper forwards the transcript

  openai/openai-agents-python#5595 · Filter by author yifanxiong272 opened on Nov 7, 2026
  1

• RunState.from_string()/from_json() replace approval-checkpoint validation errors with RuntimeError('Error details are redacted.')

  openai/openai-agents-python#5594 · Filter by author yifanxiong272 opened on Nov 7, 2026
  1

• ToolContext.approve_tool() silently drops the decision for a nested agent-as-tool approval (regression from #4413)

  openai/openai-agents-python#5593 · Filter by author yifanxiong272 opened on Nov 7, 2026
  3

• RunState.to_string()/from_string() turns assistant input messages into items the Chat Completions converter rejects

  openai/openai-agents-python#5592 · Filter by author yifanxiong272 opened on Nov 7, 2026
  3

• Feature: detect when an MCP server changes an approved tool definition

  enhancement
  openai/openai-agents-python#5588 · Filter by author emmanuelorta opened on Nov 7, 2026
  1

• Default tool error text tells the model to "try again" after a call that may already have taken effect

  openai/openai-agents-python#5573 · Filter by author astrogilda opened on Nov 7, 2026
  3

• Already-completed sibling input guardrail result omitted from tripwire run_data when the tripping result is consumed first

  openai/openai-agents-python#5567 · Filter by author lucafranco1994pot-cloud opened on Nov 6, 2026
  4

• AGENTS.md is over Codex's default 32 KiB project-doc budget

  openai/openai-agents-python#5550 · Filter by author PranavMishra28 opened on Nov 6, 2026
  3

• GenerationSpanData has no tools for Chat Completions / LiteLLM models

  enhancement
  openai/openai-agents-python#5545 · Filter by author zhoucy-syn opened on Nov 5, 2026
  4

• test_cloudflare_mount_rejects_read_only_extra_path_grant fails on Windows: expected path is hard-coded POSIX

  openai/openai-agents-python#5529 · Filter by author Mahdi-Hasan opened on Nov 5, 2026

• SessionSettings(limit=N) window can start inside a reasoning response, so the Responses API rejects the turn

  openai/openai-agents-python#5507 · Filter by author jaeminbird opened on Nov 4, 2026
  6

• List Sprites as a hosted sandbox platform

  openai/openai-agents-python#5498 · Filter by author mattp-fly opened on Nov 4, 2026

• Examples read and write text with the locale encoding, so the usaspending example test fails on non-UTF-8 Windows

  openai/openai-agents-python#5493 · Filter by author Aschenbath opened on Nov 3, 2026
  3

• Traces for runs in eu data residency

  question
  openai/openai-agents-python#5492 · Filter by author MMarciniak103 opened on Nov 3, 2026
  1

• Scope question: output-verification layers vs. agent-run guardrails, and the path to consumer flows

  openai/openai-agents-python#5490 · Filter by author PHM-Audit opened on Nov 3, 2026
  2

• Preserve wire aliases when converting Responses output items to input

  openai/openai-agents-python#5477 · Filter by author jimbobbennett opened on Nov 2, 2026
  2

• Voice: default sentence splitter never splits Chinese/Japanese, so TTS waits for the whole reply

  openai/openai-agents-python#5475 · Filter by author jaeminbird opened on Nov 2, 2026
  2

• MCPServerStreamableHttp serializes all requests on mcp 2.x, so parallel tool calls run one at a time

  openai/openai-agents-python#5474 · Filter by author jaeminbird opened on Nov 2, 2026
  5

• AnyLLMModel chat completions: extra_headers TypeError on Mistral (Chat.complete_async)

  openai/openai-agents-python#5473 · Filter by author TehilaTheStudent opened on Nov 2, 2026
  2

• tool_choice naming a function tool called web_search, file_search, mcp, etc. selects the built-in tool instead

  openai/openai-agents-python#5472 · Filter by author adityasingh2400 opened on Nov 2, 2026
  3

• OpenAIConversationsSession and OpenAIResponsesCompactionSession ignore set_default_openai_key()

  openai/openai-agents-python#5471 · Filter by author adityasingh2400 opened on Nov 2, 2026
  2

• An approved tool's guardrail results are dropped when the resumed run continues to the next turn

  openai/openai-agents-python#5470 · Filter by author adityasingh2400 opened on Nov 2, 2026
  2

• Runner.run_streamed sends a model request after a parallel input guardrail has already tripped

  openai/openai-agents-python#5469 · Filter by author adityasingh2400 opened on Nov 2, 2026
  3

• OpenAIChatCompletionsModel sends tool_choice without tools, which the API rejects, on turns with no tools

  openai/openai-agents-python#5468 · Filter by author adityasingh2400 opened on Nov 2, 2026
  5

• Docs: tool_choice="required" can select a state-changing tool when the needed one is not shown

  openai/openai-agents-python#5461 · Filter by author SnowWhiteDad opened on Oct 31, 2026
  1