Voice-Claude Assistant
A push-to-talk voice interface for Claude running as a Windows system tray app. Speak naturally, hear responses back, and when you ask Claude to *do* something it routes the task through Claude Code to Gmail, Slack, Jira, Salesforce, and Google Calendar.
- Timeline
- 2 weeks to working prototype, ongoing refinement
- Role
- Full-stack build: audio pipeline, Claude integration, MCP routing, Windows tray app
- Outcome
- Hands-free workflow access
What wasn't working.
Switching between tools all day — email, Slack, Jira, Calendar, Salesforce — was burning 15 minutes an hour on context switching. A chatbot solves some of that but still requires opening a window and typing. What if you could just speak, and have the right thing happen in the right tool, without ever stopping what you were doing?
How we built it.
A Windows system tray app that listens for F9, captures voice via Deepgram, sends the transcript to Claude for intent understanding, and routes action requests through Claude Code with full MCP tool access. Responses come back as synthesized voice through ElevenLabs. The tray icon color-codes the state so you know at a glance whether it's listening, thinking, or speaking. No terminal window, no UI to open — just hold F9 and talk.
What shipped.
- Hands-free access to Gmail, Slack, Jira, Salesforce, and Google Calendar
- Typical interaction: F9 → "Check my calendar for tomorrow and send the prep doc to the team" → 4.2 seconds later it's done
- Eliminates context-switching overhead for short, frequent tasks
- Runs quietly in the background as a tray app — zero visual footprint when idle
- Extensible via MCP — adding new tools takes minutes
Have a system like this to build?
Every project here started as a 30-minute conversation. Tell us what you're trying to ship, and we'll tell you how we'd build it.
Multiply Your Output