Claude-Code v2.1.74 Release: Memory Leak Fixes, Context Optimization, and Plugin Improvements

What's New in Claude-Code v2.1.74
Claude-Code v2.1.74 addresses several critical issues while adding new configuration options and usability improvements. This release focuses on memory management, authentication fixes, and plugin system enhancements.
Key Changes and Fixes
- Added actionable suggestions to /context command — identifies context-heavy tools, memory bloat, and capacity warnings with specific optimization tips
- Added autoMemoryDirectory setting to configure a custom directory for auto-memory storage
- Fixed memory leak where streaming API response buffers were not released when the generator was terminated early, causing unbounded RSS growth on the Node.js/npm code path
- Fixed managed policy ask rules being bypassed by user allow rules or skill allowed-tools
- Fixed full model IDs (e.g., claude-opus-4-5) being silently ignored in agent frontmatter model: field and --agents JSON config — agents now accept the same model values as --model
- Fixed MCP OAuth authentication hanging when the callback port is already in use
- Fixed MCP OAuth refresh never prompting for re-auth after the refresh token expires, for OAuth servers that return errors with HTTP 200 (e.g. Slack)
- Fixed voice mode silently failing on the macOS native binary for users whose terminal had never been granted microphone permission — the binary now includes the audio-input entitlement so macOS prompts correctly
- Fixed SessionEnd hooks being killed after 1.5 s on exit regardless of hook.timeout — now configurable via CLAUDE_CODE_SESSIONEND_HOOKS_TIMEOUT_MS
- Fixed /plugin install failing inside the REPL for marketplace plugins with local sources
- Fixed marketplace update not syncing git submodules — plugin sources in submodules no longer break after update
- Fixed unknown slash commands with arguments silently dropping input — now shows your input as a warning
- Fixed Hebrew, Arabic, and other RTL text not rendering correctly in Windows Terminal, conhost, and VS Code integrated terminal
- Fixed LSP servers not working on Windows due to malformed file URIs
- Changed --plugin-dir so local dev copies now override installed marketplace plugins with the same name (unless that plugin is force-enabled by managed settings)
- [VSCode] Fixed delete button not working for Untitled sessions
- [VSCode] Improved scroll wheel responsiveness in the integrated terminal with terminal-aware acceleration
This release is particularly important for developers working with streaming API responses on Node.js, as the memory leak fix addresses a significant resource consumption issue. The context optimization suggestions and plugin system improvements enhance daily workflow efficiency.
📖 Read the full source: GitHub Claude-Code
👀 See Also

llama.cpp Q8_0 quantization gets 3.1x speedup on Intel Arc GPUs with SYCL reorder fix
A fix to llama.cpp's SYCL backend brings Q8_0 quantization on Intel Arc GPUs from 21% to 66% of theoretical memory bandwidth, achieving 15.24 tokens/second versus 4.88 tokens/second previously on an Arc Pro B70 with Qwen3.5-27B.

Hybrid AI Architecture: Open-Source Components with Proprietary Reasoning Models
A practical hybrid AI architecture is emerging where 89% of organizations use open-source components to reduce costs by over 50%, while proprietary models handle complex reasoning tasks. Open-source frameworks offer transparency and fine-tuning capabilities without licensing negotiations.

Anthropic Reports Evidence of Mass Claude Distillation by AI Competitors
Anthropic has shared evidence that DeepSeek, Moonshot, and MiniMax were using approximately 24,000 fake accounts to conduct mass distillation of Claude, with over 16 million exchanges recorded.

EU Forces Google to Open Android AI to Third Parties Under DMA
European Commission proposes measures to allow third-party AI assistants system-level access on Android, including hot word invocation, screen context, and local model hardware access. Google calls it 'unwarranted intervention'.