News · 9 Nov 2026

Monday roundup: 9 Nov 2026

Claude Code 2.1.319 changes how auto mode behaves behind a gateway or with telemetry off, Cursor adds realtime voice calls with agents, and Google deprecates gemini-embedding-2-preview in favour of gemini-embedding-2.

  1. Anthropic

    Claude Code 2.1.319

    Released 6 Nov 2026, 17:04 UK

    Auto mode behind an LLM gateway, or with telemetry or feature flags off, now asks or denies for calls the server can't check, instead of falling back to the built-in classifier. Setting CLAUDE_CODE_AUTO_MODE_SERVER=0 keeps the old behaviour. claude -p now exits 1 when auto mode refuses actions for lack of a safety check.

    The release also adds a permissions.denyInsteadOfPrompting setting, moves MCP tasks to the background, adds CLAUDE_CODE_DISABLE_FEATURE_FLAGS, and brings several Claude apps gateway additions: pooled budget conflict rules, JSON logs, a login notice and spend in /usage.

    Fixes include Bash deny and ask rules missing commands with quoted words.

    Who it affects Anyone running Claude Code through a gateway or in CI. Check your auto mode setup before upgrading, as the exit code and fallback behaviour have changed.

    Source: Claude Code 2.1.319 release ↗

  2. Cursor

    Voice Chat with Agents

    Released 6 Nov 2026

    Realtime voice calls with agents are available in the Agents Window, IDE, web and iOS, including "Hey Siri, call Cursor". You can keep talking while the agent works, and the agent gives short spoken progress notes.

    It runs on Grok Voice, and with Privacy Mode on it runs with zero data retention. It is separate from dictation.

    Who it affects Cursor users considering voice for work, who should check the zero data retention behaviour under Privacy Mode.

    Source: Cursor changelog ↗

  3. Google

    Gemini API: gemini-embedding-2-preview deprecated

    Released 6 Nov 2026

    gemini-embedding-2-preview is deprecated. Requests now route to gemini-embedding-2 with the same schema, pricing and embedding space.

    Who it affects Developers using the preview embedding model. No re-indexing is needed, but change the model string so you're not relying on the alias.

    Source: Gemini API changelog ↗

← All news