Voice Input for Claude Cowork

Claude Cowork is Anthropic's agentic AI for knowledge work. Unlike chat, where you ask one question at a time, Cowork takes on entire tasks: research, document creation, data extraction, file organization. You describe the goal, Cowork plans and executes it, and you steer along the way.

Cowork runs inside the Claude Desktop app on macOS. It accesses your local files, connects to services like Google Drive and Gmail, and can even use your computer directly with keyboard and mouse when no connector exists. It requires a paid Claude plan (Pro, Max, Team, or Enterprise).

The fastest way to start is the keyboard shortcut. Install Spokenly, press the hotkey (customizable, works system-wide), speak, and the transcribed text appears right in the Cowork input field. No configuration, no MCP, no integration steps. It works because Spokenly inserts text at the cursor position in any app on your Mac, including Claude Desktop.

This is ideal for task descriptions. Instead of typing a detailed prompt about what you want Cowork to do, just speak it. A 30-second explanation gives Cowork significantly more context than the two sentences you would normally type. The same works for mid-task feedback: when Cowork pauses and asks a question, click the input field, press your Spokenly shortcut, and speak your answer.

For a deeper integration, you can connect Spokenly to Claude Desktop via MCP. This lets Cowork call Spokenly directly when it needs your input during an autonomous task.

Spokenly runs a local MCP server that exposes a voice dictation tool called ask_user_dictation. You add a lightweight bridge script to Claude Desktop's MCP config, and it connects via stdio transport. When Cowork needs your input, it calls this tool instead of showing a plain text prompt. A recording overlay appears in Spokenly with the agent's question. You read it, speak your answer naturally, and press Enter. The transcribed text goes back to Cowork automatically.

The difference from the hotkey approach: with MCP, Cowork initiates the voice interaction. You don't need to click the input field or press a shortcut. The agent asks, you answer, and the workflow continues. This is especially useful when Cowork asks multiple follow-up questions in a row.

Research synthesis: You ask Cowork to analyze 20 customer interview transcripts and pull out recurring themes. Midway, Cowork asks "Should I group feedback by feature area or by customer segment?" You speak your preference in five seconds instead of switching context to type.

Document preparation: Cowork is drafting a quarterly business review from your spreadsheets and notes. It asks "The Q1 revenue figures in the spreadsheet differ from the slide deck. Which source should I use?" You explain the discrepancy by voice, giving full context that helps Cowork make the right call.

File organization: Cowork is sorting hundreds of files from your Downloads folder. It asks "I found 15 files that could be either personal or work-related. Want me to show you each one, or should I use folder hints?" A quick spoken answer keeps the task moving without breaking your flow.

Data extraction: Cowork is parsing contracts into a structured spreadsheet. It asks "Some contracts list pricing per unit and others list total cost. Should I normalize everything to per-unit pricing?" You explain the business logic by voice in 15 seconds. Typing that same explanation would take two minutes.