Claude Code v2.1.195: Hook Matcher Fix, Mouse Disable Env Var, Voice Dictation Fixes
What's Changed in v2.1.195
This release addresses several long-standing issues and adds a new environment variable for controlling mouse interaction.
Key Fixes
- Hook matchers now exact-match: Previously, hyphenated identifiers like
code-reviewerormcp__brave-searchcould accidentally substring-match. Now they require exact match. To match all tools from a hyphenated MCP server, usemcp__brave-search__.*. - Voice dictation: Fixed two bugs — on macOS, long-running sessions could capture silence after the default input device changes; auto-submit never fired for languages without spaces (Japanese, Chinese, Thai).
- Plugin behavior: External plugins enabled via
.claude/settings.jsonno longer require explicit install consent on every loader path. Also fixed/plugin Enable/Disablewhen a plugin'splugin.jsonname differs from its marketplace entry name. - Background jobs: Fixed background jobs disappearing from
claude agentsor losing data when written by a newer Claude Code version. Also fixed reopening a crashed background task showing a blank screen for up to 5 seconds, and background agent daemons becoming unreachable when the control socket fails to start. - Remote session: Added a provisioning checklist display while the container starts.
Improvements
- New env var:
CLAUDE_CODE_DISABLE_MOUSE_CLICKSdisables mouse click/drag/hover in fullscreen mode while keeping wheel scroll. - Voice mode on Linux: Now distinguishes 'no microphone' from 'SoX not installed' when SoX is present but no audio capture device exists.
- claude agents UI: The completed list now fills available vertical space. On short terminals, the header compacts so live sessions stay visible.
📖 Read the full source: GitHub Claude-Code
👀 See Also
Cursor AI Study: Short-Term Speed Gains Lead to Long-Term Complexity
A study using difference-in-differences analysis found Cursor AI adoption leads to statistically significant but transient velocity increases, along with substantial and persistent increases in static analysis warnings and code complexity that drive long-term slowdowns.
OpenAI Developing GitHub Alternative According to Reuters Report
Reuters reports OpenAI is developing an alternative to Microsoft's GitHub, with the story generating 35 points and 12 comments on Hacker News.
Claude Opus 4.6 and Sonnet 4.6 now offer 1M context at standard pricing
Claude Opus 4.6 and Sonnet 4.6 now include a full 1M context window at standard pricing with no long-context premium, plus expanded media limits to 600 images or PDF pages per request.
YC-Bench Benchmark Tests LLMs as Startup CEOs, GLM-5 Shows Strong Cost-Efficiency
Researchers created YC-Bench, a benchmark where LLMs act as CEOs of simulated startups over a year, managing employees, contracts, and payroll. GLM-5 achieved $1.21M average final funds at $7.62 per run, performing within 5% of Claude Opus 4.6 which cost $86 per run.