Claude-Code v2.1.80 adds rate limit monitoring, plugin improvements, and memory optimizations
What's new in Claude-Code v2.1.80
Claude-Code v2.1.80 includes several practical improvements for developers using AI coding agents, focusing on monitoring, plugin management, and performance optimizations.
Key features and improvements
- Rate limit monitoring: Added
rate_limitsfield to statusline scripts for displaying Claude.ai rate limit usage with 5-hour and 7-day windows showingused_percentageandresets_at - Plugin marketplace source: Added
source: 'settings'option to declare plugin entries inline insettings.json - CLI tool detection: Added CLI tool usage detection to plugin tips, complementing existing file pattern matching
- Effort level control: Added
effortfrontmatter support for skills and slash commands to override the model effort level when invoked - MCP server channels: Added
--channelsflag (research preview) allowing MCP servers to push messages into your session - Memory optimization: Reduced memory usage on startup in large repositories, saving approximately 80 MB on repositories with 250,000 files
Bug fixes
- Fixed
--resumedropping parallel tool results - sessions with parallel tool calls now restore alltool_use/tool_resultpairs instead of showing[Tool result missing]placeholders - Fixed voice mode WebSocket failures caused by Cloudflare bot detection on non-browser TLS fingerprints
- Fixed 400 errors when using fine-grained tool streaming through API proxies, Bedrock, or Vertex
- Fixed
/remote-controlappearing for gateway and third-party provider deployments where it cannot function - Fixed
/sandboxtab switching not responding to Tab or arrow keys - Fixed managed settings (
enabledPlugins,permissions.defaultMode, policy-set env vars) not being applied at startup whenremote-settings.jsonwas cached from a prior session
UI improvements
- Improved responsiveness of
@file autocomplete in large git repositories - Improved
/effortto show what auto currently resolves to, matching the status bar indicator - Improved
/permissions- Tab and arrow keys now switch tabs from within a list - Improved background tasks panel - left arrow now closes from the list view
- Simplified plugin install tips to use a single
/plugin installcommand instead of a two-step flow
📖 Read the full source: GitHub Claude-Code
👀 See Also
Caveman vs 'be brief' prompt: benchmarking compression prompts for Claude
A 24-prompt benchmark across 5 arms finds that the 2-word prompt 'be brief.' matches caveman compression on both token count and output quality, though caveman provides structural consistency and safety escape features.
Opus 4.7 Token Efficiency: German Prompts Burn Up to 2x Tokens vs English
A Claude Pro subscriber reports that using German with Opus 4.7 consumed 100% of session tokens in seconds, while English used 37%. The tokenizer inefficiency stems from compound nouns and umlauts, causing 1.5–2x token usage.
SPLICE Benchmark Reveals VLMs Struggle with Temporal Reasoning, Rely on Language Priors
Research presented at EMNLP 2025 shows vision-language models score poorly on a video sequencing task where humans excel, with models like Gemini 2.0 Flash reaching 51% accuracy versus human performance of 85%. Models frequently rely on visual shortcuts and language descriptions rather than true visual understanding.
Fine-tuning Phi-4-mini by training only LayerNorm parameters fails to improve performance
A hobbyist tested training only LayerNorm γ values on Phi-4-mini across Python and medical domains with different learning rates and data formats. Performance degraded slightly on all benchmarks compared to baseline, with the author concluding transformers already route information dynamically through attention.