NextClaw v0.49.0

Released: 2026-09-09

Tags: voice input, CLI, models, sessions, mobile

NextClaw v0.49.0 adds controlled voice input to chat, object-level CLI commands for providers, models, and web search, and more reliable loading, recovery, and tool history for large sessions.

New features

• Chat now supports click-to-record voice input, with an optional push-to-talk shortcut on desktop. Transcribed text appears at the current cursor position and remains in the draft when recording ends; it is never sent automatically.
• The nextclaw CLI can now add and manage model providers, discover and configure models, test connections, authorize supported providers, select the default model, and configure web search.

Improvements

• Session lists and large conversations load faster without refreshing the entire list for every live status change.
• The model picker puts favorites and recently used models first, then groups the remaining models by provider.
• Mobile chat hides the avatar beside your own messages to leave more room for text. Returning from the background or reconnecting after a network interruption now replaces stale real-time connections automatically.

Fixes

• Chat no longer remains stuck on “Thinking” after a task has already finished.
• Refreshing a conversation no longer duplicates multi-step tool messages or leaves an endless waiting indicator.
• Successful command cards now show terminal output such as stdout and stderr.
• Completed messages preserve and display their elapsed time, including after a page refresh.

What you need to do

Update normally. The first time you use voice input, allow microphone access when prompted. The desktop push-to-talk shortcut remains optional.