◉CLAUDE LABJP
●FORK — Claude Code 2.1.212 changes what /fork does: it copies your conversation into a new background session with its own row in claude agents, so you can keep working. The old in-session subagent is now /subtask●LIMITS — WebSearch calls are now capped at 200 per session by default, and subagent spawns get the same 200 ceiling, so a runaway search or delegation loop stops on its own●MCPBG — MCP tool calls running past two minutes now move to the background automatically, keeping the session usable. Tune the threshold with CLAUDE_CODE_MCP_AUTO_BACKGROUND_MS●PLANFIX — Fixed plan mode auto-running file-modifying Bash commands such as touch and rm without a permission prompt or an SDK canUseTool callback●SONNET5 — Claude Sonnet 5 is running on introductory pricing of $2 per million input tokens and $10 per million output. After August 31 it moves to $3 and $15●IPO — Bankers are reportedly lining up investor meetings for Anthropic ahead of a possible public listing as soon as October●FORK — Claude Code 2.1.212 changes what /fork does: it copies your conversation into a new background session with its own row in claude agents, so you can keep working. The old in-session subagent is now /subtask●LIMITS — WebSearch calls are now capped at 200 per session by default, and subagent spawns get the same 200 ceiling, so a runaway search or delegation loop stops on its own●MCPBG — MCP tool calls running past two minutes now move to the background automatically, keeping the session usable. Tune the threshold with CLAUDE_CODE_MCP_AUTO_BACKGROUND_MS●PLANFIX — Fixed plan mode auto-running file-modifying Bash commands such as touch and rm without a permission prompt or an SDK canUseTool callback●SONNET5 — Claude Sonnet 5 is running on introductory pricing of $2 per million input tokens and $10 per million output. After August 31 it moves to $3 and $15●IPO — Bankers are reportedly lining up investor meetings for Anthropic ahead of a possible public listing as soon as October
← Back to Blog

Claude Code Voice Mode: The Complete Guide to Hands-Free Coding

Claude CodeVoice ModeHands-FreeProductivityNew Feature

What Is Voice Mode?

In March 2026, Claude Code introduced voice mode — a feature that lets you code by speaking. Simply type /voice to activate it, and you can interact with Claude using natural speech while keeping your hands free.

The key design decision behind voice mode is its push-to-talk approach. You hold down the spacebar while speaking and release it to send your command. The microphone is never passively listening, which means no accidental commands and no privacy concerns about your terminal eavesdropping on conversations.

Getting Started

1. Grant Microphone Access

The first time you use voice mode, your operating system will ask for microphone permission. On macOS, go to System Settings → Privacy & Security → Microphone, then enable access for your terminal app (Terminal, iTerm2, Warp, etc.).

2. Activate Voice Mode

In any Claude Code session, type:

/voice

That's it. Run /voice again to toggle it off.

3. The Basic Workflow

The interaction flow is straightforward:

  1. Hold the spacebar and speak your command
  2. Release the spacebar to send the audio
  3. Claude displays the transcript in your terminal
  4. Review the transcript, then Claude executes the task

Since the transcript appears before execution, you can catch any misheard words and cancel before Claude acts on them.

Practical Use Cases

Code Reviews

Verbally point Claude to specific files or functions for review:

"Check the useEffect in src/components/UserProfile.tsx
  and see if there are any missing dependencies"

Refactoring Instructions

Describe complex refactoring goals in natural language:

"Split this function into three smaller helpers —
  one for validation, one for data transformation,
  and one for the API call"

Debugging Conversations

Explain the bug you're seeing and let Claude investigate:

"The build passes but after login there's an infinite
  redirect loop. Can you trace the auth flow?"

Key Things to Know

Seamless Text and Voice Switching

You can switch between voice and text input at any time without losing context. This lets you use voice for high-level directions and text for precise code snippets or special characters — the best of both worlds.

Free Transcription Tokens

Transcription tokens are not counted against your rate limits. You can use voice as much as you want without impacting your regular usage quota.

20 Languages Supported

As of March 2026, voice mode supports 20 languages, including Japanese, Spanish, French, German, and more. Cross-lingual workflows are fully supported — you can give instructions in one language and generate code or documentation in another.

Tips for Getting the Most Out of Voice Mode

Be specific with your instructions: Just like with text prompts, clarity matters. State what you want, where, and how for the best results.

Use short phrases: Breaking your instructions into shorter phrases tends to produce more accurate transcriptions than long, unbroken sentences.

Know when to type: Regular expressions, special symbols, and precise variable names are often easier to type. Let voice handle the natural language parts.

Use a good microphone: A noise-canceling microphone significantly improves recognition accuracy in noisy environments.

Wrapping Up

Voice mode in Claude Code opens up a new dimension of developer productivity. Being able to keep your train of thought flowing without pausing to type can be a game-changer, especially during code reviews, refactoring sessions, and debugging conversations.

The feature is currently in a gradual rollout — if /voice is available in your Claude Code session, give it a try. You might be surprised how natural it feels to talk your way through a codebase.