Claude-Code v2.1.30 Released with PDF and OAuth Enhancements
Claude-Code's latest release, v2.1.30, brings notable enhancements and fixes that developers should be aware of. This version introduces a new parameter for the Read tool, allowing specific page ranges to be extracted from PDFs, like pages: "1-5". Additionally, handling of large PDFs (over 10 pages) now returns a lightweight reference instead of embedding the content directly, improving context efficiency when mentioned.
For developers integrating with MCP servers that lack Dynamic Client Registration, the update offers pre-configured OAuth client credentials. Utilize the --client-id and --client-secret options when adding with claude mcp add.
The release also includes updates for debugging with the new /debug command, support for additional git log and git show flags in read-only mode, such as --topo-order and --cherry-pick, and improved visualization in VSCode with multiline input support for question dialogs (Shift+Enter for new lines).
Key bug fixes in this release address phantom "(no content)" text in API history, prompt cache invalidation issues, and 400 errors after using /login. Additionally, there's a 68% reduction in memory usage for sessions thanks to optimized stat-based loading.
This update is valuable for developers looking to efficiently handle large document processing and improve integration with third-party MCP services. Users experiencing issues with session handling or permissions will also find these improvements beneficial.
📖 Read the full source: GitHub Claude-Code
👀 See Also
Claude Opus 4.6's effort=low parameter differs from other providers' low-reasoning modes
Claude Opus 4.6's effort=low parameter controls general behavioral effort, not just reasoning depth, unlike OpenAI's reasoning.effort=low or Gemini's thinking_level=low. This caused agents to make fewer tool calls, be less thorough in cross-referencing, and ignore parts of system prompts about web research.
GPT-5.5 on OpenClaw: Users Report 5-Hour Limit Burned in 3 Prompts, Broken Code
Reddit user reports GPT-5.5 on OpenClaw consumed 5-hour limit in 3 prompts, introduced bugs; DeepSeek V4 Pro fixed the mess.
OpenAI disbands preparedness team weeks before critical model slowdown
OpenAI shut down its preparedness team in July, weeks before a model hit a critical cyber threshold. The team that assessed catastrophic risks no longer exists as a unit.
Claude Prompt Cache Diagnostics: Stats Thread Reveals 98.9% Cache Read Ratio
Two days ago, Claude released prompt cache diagnostics in Console. One developer reports 98.9% cache read ratio, with 80% of misses due to messages changed.