Speak your prompts.Code faster on Windows.
Stop typing every prompt twice. Hold a hotkey, speak it, and watch it land in Claude Code, Codex CLI, or Cursor. The minutes you save on each prompt compound into hours you keep for actual work.
Local transcription, private by default, no account required.
10 free uses, then a 7-day trial — no credit card required.
Save typing minutes
Hold a hotkey and dictate. PromptPaste pastes the text where your cursor is, faster than you can type it.Built for your AI tools
Lands in Claude Code, Codex CLI, Cursor, and anywhere else you talk to an AI on Windows.Skip the file picker
Paste clipboard screenshots straight into supported apps — no save dialog, no rename, no detour.Audio stays local
Whisper runs on your device. Nothing you say or paste leaves your machine.Hold your flow
Less tab-switching, less keyboard friction, less context-switching when you're deep in a problem.Draft with voice. Attach with paste. Stay in flow.
PromptPaste stays voice-first for Windows developers using Claude Code, Codex CLI, and Cursor. When a bug report, mockup, terminal error, or diagram is easier to show than describe, paste the screenshot directly into supported apps instead of stopping to hunt through a file dialog.
Speak the first draft
Draft prompts for Claude Code, Codex CLI, or any terminal by voice while the cursor stays where you need it.
Paste screenshots where supported
Copy terminal captures, UI mockups, or error screenshots to the clipboard and paste them into apps that accept pasted images.
Voice plus screenshot context
Combine voice-drafted prompts with screenshot context for faster bug triage and implementation handoffs in Claude Code, Codex CLI, and Cursor.
Start with the workflows PromptPaste is built for
Voice input is the core workflow for Claude Code, Codex CLI, and Windows Terminal. Screenshot paste extends that loop when the task needs visual context in a supported app.
How it works
PromptPaste is a Windows input loop for Claude Code, Codex CLI, and terminal workflows: voice stays primary, screenshot paste adds visual context where the target app accepts pasted clipboard images.
Focus the target workflow
Put focus on the terminal, AI composer, or app input where the next draft or attachment should land.
You stay inside the Windows workflow you are already using.
Speak the first draft
Hold the configured hotkey and say the prompt, instruction, command, or notes in one natural pass.
Voice handles the slowest part of drafting without forcing a new writing surface.
Copy an image when the task needs context
If a screenshot, mockup, terminal error, or diagram explains the task faster, copy it to the Windows clipboard.
You can add visual context without dropping into a file picker.
PromptPaste keeps voice local
Voice transcription stays on-device on Windows. For supported apps, PromptPaste can paste the clipboard image into the focused composer.
The flow stays fast, local-first, and close to the cursor.
Review and send
Check the drafted text, confirm the image landed where expected, then send both into the workflow.
You move from idea to prompt faster with less friction.
See PromptPaste in a Claude Code and Codex CLI workflow
This is the core loop: press the hotkey, speak the prompt, watch PromptPaste transcribe locally, and keep moving in Claude Code, Codex CLI, or Windows Terminal.
Works with Claude Code and Codex CLI on Windows.
Inserts text directly at the terminal cursor.
Transcribes locally on-device with no cloud speech pipeline.
Fits Windows Terminal and other focused terminal inputs.
Built for Windows developers who want terminal speed in Claude Code and Codex CLI without opening another app or browser tab.
More Windows workflows once the input loop clicks
PromptPaste also fits git writing, kubectl triage, PowerShell runbooks, SQL CLI work, release validation, incident notes, and screenshot-based prompting where pasted images are supported.
Privacy first
PromptPaste is local-first by design. Transcription runs on-device - privacy-sensitive workflow data is never pushed to third-party AI endpoints.
On-device speech
Audio and transcription stay local during normal use.
No cloud transcript store
Prompts are not uploaded to external storage services by default.
Minimal auth surface
Billing/auth checks are separated from voice transcription payloads.
User control
You can review privacy docs and support policies before subscribing.
Frequently asked questions
Practical answers about setup, privacy, compatibility, image paste, and how PromptPaste fits Windows AI workflows.
Yes. Transcription runs entirely on your machine using an on-device model. No internet connection is required during dictation.
No. Audio is processed locally in transient buffers and is never stored to disk or uploaded to any server.
PromptPaste targets the active or locked window and inserts text at your cursor position using standard input simulation. It is built for cursor-level terminal input, not browser-based dictation.
PromptPaste uses push-to-talk: hold the configured hotkey, speak your prompt or command, and release. Text is inserted at the cursor when you let go.
Yes. PromptPaste can paste clipboard images into supported apps and workflows on Windows. Voice stays the primary input flow, and image paste is there when a screenshot or mockup explains the task faster.
It works where the target app already accepts pasted clipboard images in the focused composer or input area. Typical examples include Claude, Codex, ChatGPT desktop or web, and similar AI tools that support screenshot pasting.
PromptPaste uses the local Windows clipboard for image paste. PromptPaste does not need to upload the image itself, but the receiving app handles pasted content under its own privacy and retention rules.
Yes. A common flow is to draft the request by voice, copy a screenshot or mockup, paste it into the target app, then send both together.
The most common causes are a text-only field, the wrong element being focused, or an app that does not accept pasted clipboard images in that workflow. If normal image paste does not work there, PromptPaste cannot force support.
Windows Terminal, PowerShell, cmd, VS Code integrated terminal, and most terminal emulators with editable input focus are supported.
Yes. PromptPaste is specifically positioned for Claude Code and Codex CLI workflows on Windows - speak the prompt, review it at the cursor, and run.
14 languages: English, French, Spanish, German, Italian, Portuguese, Dutch, Japanese, Korean, Chinese, Russian, Arabic, Hindi, and Hebrew. The right Whisper model loads automatically — an English-optimized model for English and a multilingual model for everything else (auto-downloads ~140 MB once on first non-English use, then cached). Pick your language in Settings → Language.
Yes. PromptPaste can be deployed in air-gapped and restricted networks with a lifetime enterprise license. We build custom packages for your environment with offline activation and no internet dependency. Visit /enterprise or contact us for details.
Install PromptPaste — the voice layer for AI coding on Windows
Local voice transcription, cursor-level text insertion, and screenshot paste for supported apps. Built for Windows developers who use terminal-first AI workflows without breaking focus.