Who contributed to AI Agents on August 5, 2026?

13 developers shipped this update, including imnishitha, dependabot, neubig, all-hands-bot, hieptl, sleeyax, oab24413gmai, and abeatrix, and 5 more.

What were the notable AI Agents updates?

fix(langchain): stop HITL approval gates from silently failing open, chore: bump aiohttp from 3.14.1 to 3.14.3 in /libs/langchain, and chore: bump aiohttp from 3.14.1 to 3.14.3 in /libs/partners/xai.

AI Agents

LangChain, OpenHands, Cline, and LlamaIndex - agent frameworks and coding agents

Pick a date

Topics: AI Agents AI / ML Full archive →

The Wire · Showcase

HITL APPROVAL GATES STOP SILENTLY FAILING OPEN, AIOHTTP FIXES REDIRECT LEAKS ACROSS LANGCHAIN STACK

By RepoJournal · Filed 07:02 UTC on August 5, 2026 · About AI Agents

13 people shipped this

dependabot @dependabot 4 cited

abeatrix @abeatrix 2 cited

shivamlalakiya @shivamlalakiya 2 cited

imnishitha @imnishitha 1 cited

neubig @neubig 1 cited

all-hands-bot @all-hands-bot 1 cited

gkorland @gkorland 1 cited

hieptl @hieptl

sleeyax @sleeyax

oab24413gmai @oab24413gmai

mkondratek @mkondratek

Anai-Guo @Anai-Guo

Asjad Abbas @asjad3

LangChain plugged a critical security hole in human-in-the-loop middleware where invalid configurations silently disabled approval gates, while simultaneously patching a cascade of aiohttp vulnerabilities that leak headers across origin boundaries.

The HumanInTheLoopMiddleware now raises ValueError immediately if allowed_decisions config is missing, empty, or misspelled [1], stopping approval gates from silently defaulting to open. This was a silent failure mode that could allow unauthorized tool execution in production agents. In parallel, LangChain bumped aiohttp from 3.14.1 to 3.14.3 across the core library [2], the XAI partners module [3], the HuggingFace integration [4], and the Fireworks integration [5], fixing a client bug where only the first Authorization, Cookie, and Proxy-Authorization headers were dropped on cross-origin redirects. That fix matters: agents making authenticated requests through proxies or redirect chains were leaking credentials. OpenHands shipped MCP credential preservation during Canvas mutations [6] and tightened CI enforcement requiring frontend screenshot/video evidence and human sign-off [7]. Cline added lazy session persistence and initiation mode to the SDK [9], while fixing connector instance claiming before socket connect [8]. LLama Index fixed FalkorDB's schema, vector ranking, and Cypher injection vulnerabilities in property graph retrieval [10], corrected Weaviate metadata for pre-existing collections [11], and opened Memory to any AsyncDBChatStore instead of just SQLAlchemy [12].

Action items

→ Upgrade to the latest LangChain with aiohttp 3.14.3 before next deploy if using authenticated proxy or redirect chains langchain-ai/langchain [immediate]
→ Review HumanInTheLoopMiddleware configs for empty or misspelled allowed_decisions to catch misconfigurations now langchain-ai/langchain [plan]
→ Test FalkorDB property graph stores against real containers if using TextToCypherRetriever or VectorContextRetriever run-llama/llama_index [plan]

References

[1] fix(langchain): stop HITL approval gates from silently failing open ↗ langchain-ai/langchain
[2] chore: bump aiohttp from 3.14.1 to 3.14.3 in /libs/langchain ↗ langchain-ai/langchain
[3] chore: bump aiohttp from 3.14.1 to 3.14.3 in /libs/partners/xai ↗ langchain-ai/langchain
[4] chore: bump aiohttp from 3.14.1 to 3.14.3 in /libs/partners/huggingface ↗ langchain-ai/langchain
[5] chore: bump aiohttp from 3.14.1 to 3.14.3 in /libs/partners/fireworks ↗ langchain-ai/langchain
[6] fix: preserve MCP credentials during Canvas mutations ↗ OpenHands/OpenHands
[7] ci: require frontend screenshot/video and checked human-tested box ↗ OpenHands/OpenHands
[8] fix(cli): claim connector instance before socket connect ↗ cline/cline
[9] feat(sdk): add session initiation mode and lazy session persistence ↗ cline/cline
[10] fix(falkordb): correct schema, vector ranking, filters and batching in property graph store ↗ run-llama/llama_index
[11] fix(weaviate): return real properties as node metadata for pre-existing collections ↗ run-llama/llama_index
[12] feat(core): allow Memory to accept any AsyncDBChatStore ↗ run-llama/llama_index

Quick answers

What shipped in AI Agents on August 5, 2026?: LangChain plugged a critical security hole in human-in-the-loop middleware where invalid configurations silently disabled approval gates, while simultaneously patching a cascade of aiohttp vulnerabilities that leak headers across origin boundaries. In total, 58 commits and 63 pull requests landed.
Who contributed to AI Agents on August 5, 2026?: 13 developers shipped this update, including imnishitha, dependabot, neubig, all-hands-bot, hieptl, sleeyax, oab24413gmai, and abeatrix, and 5 more.
What were the notable AI Agents updates?: fix(langchain): stop HITL approval gates from silently failing open, chore: bump aiohttp from 3.14.1 to 3.14.3 in /libs/langchain, and chore: bump aiohttp from 3.14.1 to 3.14.3 in /libs/partners/xai.

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

AI Agents Aug 4

LANGCHAIN PLUGS RESOURCE LEAK, CLINE ADDS CONTEXT-WINDOW RECOVERY

LangChain fixed a critical event-loop leak in streaming tracers that was breaking Python 3.14+ tests, while Cline shipped context-window overflow detection and recovery across its SDK.

+10

ai-agents 73 shipped

AI Agents Aug 3

LANGCHAIN LOCKS DOWN TOOL ARGUMENT HANDLING, CLINE HARDENS RELEASE PIPELINE

LangChain shipped a burst of tool-handling fixes overnight that close injection gaps in subclasses, while Cline moved desktop signing secrets behind gated environments and killed notification spam.

ai-agents 26 shipped

AI Agents Aug 2

CLINE CHECKPOINTS NOW REWIND RELIABLY ACROSS CLI AND VSCODE

Cline shipped a four-commit fix that makes workspace snapshots actually work as true rewinding across both the VS Code extension and CLI after days of checkpoint failures broke both surfaces.

ai-agents 40 shipped

Elsewhere on the wire

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.

+10

llms 240 shipped

@CachyOS Aug 6

WIKI TRANSLATION ROLLBACK CHAOS, MIRROR STACK HOLDING STEADY

CachyOS wiki maintainers are caught in a revert loop over German translations while the status infrastructure logs 55 commits without incident.

infra 57 shipped

Elixir & Phoenix Aug 6

ELIXIR BUMPS ZIZMOR SECURITY SCANNER TO 1.28.0

The Elixir repo updated its CI security action overnight, pulling in the latest zizmor static analysis tooling.

elixir 1 shipped

@dotnet Aug 6

DOTNET FOUNDATION PULLS iOS METRICS INTO FOCUS AS RUNTIME HARDENS HTTP PARSING

Performance team fixed iOS size reporting to measure only distributable packages, while runtime clamped down on RFC 9112 compliance for chunked encoding.