Who contributed to OpenAI on August 4, 2026?

8 developers shipped this update, including thomas, rhan-oai, jif, copyberry, Hayden, Justin Beckwith, github-actions[bot], and Alex Chang.

What were the notable OpenAI updates?

Add MCP client conformance regression gates (#36810), v7.4.0, and v2.53.0.

OpenAI

Codex, the SDKs, and the engine behind ChatGPT

Pick a date

Topics: AI / ML Full archive →

The Wire · Showcase

GPT-5.5 LANDS ACROSS NODE AND PYTHON SDKS, CODEX HARDENS MCP CONFORMANCE

By RepoJournal · Filed 06:58 UTC on August 4, 2026 · About OpenAI

8 people shipped this

copyberry @copyberry 2 cited

Justin Beckwith @jbeckwith-oai 2 cited

github-actions[bot] @github-actions[bot] 2 cited

Alex Chang @apcha-oai 2 cited

thomas @tsarlandie-oai 1 cited

rhan-oai @rhan-oai 1 cited

jif @jif-oai

Hayden @HAYDEN-OAI

OpenAI shipped gpt-5.5 model support and tool metadata across both JavaScript and Python SDKs while Codex locked down a full MCP client conformance regression suite to prevent protocol slippage.

The SDK team moved fast on model expansion. Both openai-node v7.4.0 [2] and openai-python v2.53.0 [3] now recognize gpt-5.5 as a valid model across chat, responses, and tool-calling surfaces, with optional name and namespace metadata fields on tool output items [4] [5]. These are backward-compatible additions that let you start targeting the new model without breaking existing code. On the infrastructure side, both Node and Python repos completed their migration off Stainless-owned CI tooling [6] [7], replacing custom release-please triggers with the upstream googleapis action and retiring legacy branch filters, runner selection, and artifact upload helpers. The Python team also fixed a dependency build issue to avoid NumPy source builds that were bloating CI runs. Over in Codex, the real infrastructure story is a conformance regression matrix that runs the Codex executable against a pinned official MCP client conformance suite across shipping, intermediate, and modern protocol versions, plus HTTP and stdio transports and OAuth scenarios [1]. Model instruction consolidation [8] removed the in-memory instruction path and standardized on model_messages for bundled, remote, fallback, and overridden metadata. Linux managed proxy helpers got hardened to detach standard I/O streams, move readiness pipes above the standard descriptor range, and treat zombie processes as exited [9]. The Codex Apps catalog limit jumped from 2,048 to 8,192 items for host-owned registrations while keeping the lower limit for third-party MCP servers [10].

Action items

→ Upgrade openai-node to v7.4.0 and openai-python to v2.53.0 if targeting gpt-5.5 openai/openai-node [plan]
→ Review Codex MCP conformance regression matrix for custom protocol versions in your deployment openai/codex [monitor]
→ Test managed proxy helper lifecycle changes on Linux deployments before rolling out openai/codex [plan]

References

[1] Add MCP client conformance regression gates (#36810) openai/codex
[2] v7.4.0 ↗ openai/openai-node
[3] v2.53.0 ↗ openai/openai-python
[4] feat(api): Add gpt-5.5 model and tool metadata fields (#2049) openai/openai-node
[5] feat(api): Add gpt-5.5 and tool name/namespace to Responses types ↗ openai/openai-python
[6] ci: retire obsolete Stainless infrastructure (#2067) openai/openai-node
[7] ci: migrate release-please to upstream action (#3568) openai/openai-python
[8] Consolidate model instructions in `ModelMessages` (#36787) openai/codex
[9] Harden Linux managed proxy helper lifecycles ↗ openai/codex
[10] Raise the host-owned Codex Apps catalog limit ↗ openai/codex

Quick answers

What shipped in OpenAI on August 4, 2026?: OpenAI shipped gpt-5.5 model support and tool metadata across both JavaScript and Python SDKs while Codex locked down a full MCP client conformance regression suite to prevent protocol slippage. In total, 51 commits, 59 pull requests, and 5 releases landed.
Who contributed to OpenAI on August 4, 2026?: 8 developers shipped this update, including thomas, rhan-oai, jif, copyberry, Hayden, Justin Beckwith, github-actions[bot], and Alex Chang.
What were the notable OpenAI updates?: Add MCP client conformance regression gates (#36810), v7.4.0, and v2.53.0.

CODEX HARDENS SKILL SELECTION AND SESSION LOOKUP, NODE SDK CLOSES LINTER GAPS

Codex decouples explicit skill selection into a dedicated crate while tightening TUI session validation, as the Node SDK completes its oxlint cleanup stack across five rule categories.

ai 118 shipped

OpenAI Aug 5

CODEX HARDENS SECURITY, SDKS MODERNIZE TOOLCHAINS

Codex now prompts users before trusting local directories, closing a prompt injection exposure, while both Node and Python SDKs shift to Oxlint and Oxfmt tooling.

ai 114 shipped

OpenAI Aug 3

CODEX HARDENS LOGIN FLOW WITH ONBOARDING HINTS AND METADATA PROTECTION

Three critical fixes shipped across auth, data persistence, and usage tracking to prevent thread metadata loss and secure OAuth callback handling.

ai 6 shipped

OpenAI Aug 2

CODEX SHIPS AGENT PLUGIN OVERHAUL AND TUI PERFORMANCE FIXES

Codex now handles portable Agent Plugins with schema-declared manifests and cuts terminal redraw overhead by caching screen dimensions instead of querying on every frame.

ai 12 shipped

Elsewhere on the wire

AI Agents Aug 6

LANGCHAIN HARDENS ANTHROPIC TOOL CALLING AND GIT CREDENTIALS AS OPENHANDS SHIPS 1.10.0

LangChain fixed a critical credential leak in evaluation runs while Anthropic tool schemas get proper fallback handling, as three major agent frameworks push releases across the ecosystem.

ai-agents 54 shipped

Local LLMs Aug 6

OLLAMA FIXES KV CACHE CORRUPTION, LLAMA.CPP HARDENS FILE SEARCH, VLLM KILLS MOE LEGACY CODE

Ollama patched a critical cache corruption bug that corrupts lazy snapshots during multi-request prefills, while llama.cpp hardened Windows junction traversal and vLLM removed 14 hardcoded MoE shards across quantization methods.