Repository navigation
Feature Request: Global Multi-Session Voice Control & Project Switcher #305
Description
I'm a heavy (and very happy) Happy user running multiple Claude Code sessions across different projects throughout the day. The voice agent is fantastic within a single session, but right now it feels "trapped" inside whichever project I'm currently viewing.
My typical workflow:
I have 3-5 active sessions going (e.g., a startup pitch deck, a podcast tooling repo, an F1 data project, and a client API)
I'm away from my desk, using voice to check in on progress
I want to ask "what's Claude doing in the podcast repo?" without manually navigating to that session first
Currently, I have to: open the app → find the session in the sidebar → tap into it → then use voice. This breaks the hands-free workflow that makes Happy so powerful.
Proposed Solution
A "global" voice mode that sits above individual sessions and can route commands across all of them.
Core capabilities:
List sessions by voice
"What sessions are running?"
"Show me my active projects"
Response: "You have 4 active sessions: F1 Data Analysis (idle 5 min), Podcast Tooling (Claude is editing files), Startup Pitch (waiting for permission), API Client (completed task)"
Switch sessions by voice
"Switch to the F1 project"
"Jump to my most recent session"
"Go to the podcast tooling repo"
Fuzzy matching on project name, folder name, or custom session names (#290)
Query sessions without switching
"What is Claude doing in the podcast repo?"
"Does the startup pitch session need my input?"
"Which sessions are waiting for permission?"
Get status updates without leaving current context
Session lifecycle via voice
"Start a new session in ~/projects/new-thing"
"Close the startup pitch session"
"Pause the API client" (if pause/resume becomes a thing)
Smart routing
If I say "commit these changes" while in global mode, Happy could ask "Which project?" or infer from context
"Tell the F1 project to run the tests" — route command to specific session
Implementation ideas:
Persistent system-level voice mode: A voice agent that runs independently of any single session, always listening (with wake word or push-to-talk)
Global session registry: Happy already knows about all sessions (they're in the sidebar). Expose this to the voice agent.
Session nicknames: Let users assign memorable names ("F1", "podcast", "client work") for easier voice targeting. Builds on #290.
Non-Voice Equivalent (Optional but Nice)
Why This Matters
Happy's superpower is untethering me from my desk while Claude works. But with multiple projects, I'm still mentally managing "which session am I in?" The voice agent should handle that cognitive load — I just want to talk to my AI coding assistant, and Happy figures out the routing.
Think of it like: today's voice agent is a walkie-talkie to one person. I'm asking for a dispatcher who can connect me to any of my team members.
Related Issues
#195 - List/attach to background sessions from terminal
#290 - Custom session name flag
#291 - Multiple devices to one computer
#272 - New session page with profiles (great foundation for this)
Thanks for building Happy!! It's already changed how I work. This would take it to the next level. 🙌
Activity
bra1nDump commentedon Jan 7, 2026
Love to see more power voice users! And thank you for this in depth analysis / feature request. I think this issue can be a good specification of the next voice agent iteration. I fully share your excitement and your final analogy spot on
At this time I don't think I can myself prioritize building this because we have other issues like stability to take care of - but I do think that voice can really be pushed much further with even current capabilities of models
jakenuts commentedon Jan 19, 2026
Absolutely, this would be a huge step forward. The voice integration is already a brilliant feature, not because of the mode it allows but the clever introduction of an intermediary, a project manager between you and the agent.
That idea to forgo a basic TTS interface and instead extend the entire workflow, I wish that I had such sparks of genius in my changelog. And it points to a future interface for agents that is likely only now occuring to alot of developers - like the "AI in your IDE", this session/project based interaction is going to be short lived and supplanted by an agent that knows the projects/tasks and can move between them handing off tasks and reporting on progress. Or a set of agents each dedicated to a particular project to preserve context and then coordinating with their boss who in turn coordinates with the dopey meat-brain.
Long, gushing story short, raising the voice assistant one level up to handle projects as well as tasks would be a revolutionary update.
hyacz commentedon Jan 26, 2026
What a brilliant idea!
jakenuts commentedon Jan 31, 2026
Wait! It's already there!?!?
Yesterday while heading out to the mountains to go gold prospecting I opened up a session in happy on my iPhone and started a voice session, and at some point I switched to a different session I wanted to check on, but the voice was still talking about the previous one. It started to instruct this new agent in how to proceed for the task in the other session and I interrupted and said "this is the wrong session" and the voice concurred and said it would send that reponse to the other session. When I asked if it could traverse sessions and interact with each of them it said "yes, of course". So it seems like either I hit an odd edge case or my understanding of the voice model being rooted in one session or the other was mistaken. I haven't experimented with it yet (all day on the river) but it makes sense that a basic function of the voice model would be session switching and awareness of which session is being discussed so it could just be baked in. ?