Bring your agent along
Keep working with your coding agent. OpenSaysMe gathers what you’re seeing and saying, so you can hand over the problem without writing it all out.
Point at the bug.
Say “fix this.”
Point at a failing test or UI bug and say what needs fixing. OpenSaysMe gathers context across your apps and can act on your desktop or pass the request to your coding agent. For macOS.
Give your coding agent
the context behind “this.”
Show the UI bug
Circle the broken layout with your mouse as you explain it. Your voice, screen, and cursor movement give “this looks wrong” a specific place to look.
Add the failing test
Switch to your terminal and point at the failure or select the stack trace. Bring the error and the visual issue together in one request to your agent.
Keep the conversation going
“It happens on mobile, too.” “Here’s the error.” Recent requests and results carry into short follow-ups, so you can keep explaining without starting over.
Stay in your flow
Use one shortcut from your browser, editor, or terminal. Talk through what you’re debugging as you move between apps. Press Escape to stop.
A little face, useful feedback
A quiet overlay shows when OpenSaysMe is listening, thinking, or acting. Keep an eye on the request without leaving the code you’re working on.
Know what your
agent can see
OpenSaysMe runs on your Mac and calls OpenAI directly with your API key. Here’s how your voice and desktop context are used.
You choose when to speak
Tap your shortcut to begin recording, then tap again to finish. Escape cancels the current request.
Your key, your connection
Bring your own OpenAI API key. Setup saves its environment-variable name, never the key itself.
Connections & context
- Your voice
api.openai.comAfter activation, microphone audio streams directly to OpenAI for transcription.
- Your screen
OpenAI reasoningScreen frames, window and selection context, cursor history, and tool results help the model understand your request.
- Your API key
Environment variableRead from your local environment when you run the app. You control the account and usage.
- Recent work
Saved on your MacApp and window names are remembered for up to eight hours and shared as context. View or clear them with
osm context.- Follow-ups
Short-lived memoryRecent requests and results help continue the conversation. This context expires after two minutes or when you cancel.
- Debug captures
Off by defaultOpt-in debugging saves screenshots, transcripts, and diagnostics locally until you delete them.
Bring OpenSaysMe into your dev workflow
Open source. Built for your development workflow. Run it, explore the code, or make it yours.
Build for macOS
macOS 14+, Node.js 26.13+, Xcode Command Line Tools, and your OpenAI API key.
Build from sourceWhat you’ll need
Grant Microphone, Accessibility, and Screen Recording permissions during setup. The README walks you through it.
Show your first bug
Walk through a UI issue, add its terminal error, and ask OpenSaysMe to tell your coding agent.
Read the guideThe first thing to try
Press Ctrl + Shift + Space, point at the bug while you talk, then show the terminal error. Ask it to tell your coding agent, and press the shortcut again to finish.
Make it yours
Explore the code, share an idea, or report something that could work better. Built in the open, under MIT.
Share an idea View on GitHubStart with the source
git clone https://github.com/andrewruba/opensaysme.gitIn the opensaysme directory
npm ci && npm run install:cli && osm setup