Invoko
A little hand on your Mac
1.3K followers
A little hand on your Mac
1.3K followers
Invoko is an AI desktop helper you can talk to while you work. Bring it beside anything on your screen, ask it questions, or let it handle tasks across your apps.
I am curious about email management. can Invoko distinguish between high priority emails and messages that can safely wait?
@daniel_harris11 Emails are always a pain in the ass cuz there're just too many! Currently we mainly use pure model capabilities to determine whether the email is of high priority or not, but we are looking for a more solid solution for that ofc. Stay tuned!
how does the agent handle situations where multiple apps are requesting attention at the same time?
@luz_bidelspach It can handle tasks in parallel so that wouldn't be a problem at all haha
what happens if Invoko misunderstands a task or schedules something incorrectly? is there an easy rollback mechanism?
@fletcher_oliver Great catch! Currently Invoko will make a confirmation first before doing that, but a rollback mechanism will def be something interesting. Noted!
how customizable are the actions and workflows? can users create their own automation rules?
@rivendell_chorro yes, just tell Invoko your preference and it'll remember that
how does Invoko compare to browser based AI agents when it comes to speed and context awareness?
@vespertine_dallarosa It will outperforms browser based AI agent for sure since it's local deployed and gets more context besides browser.
Looks like an interesting way to use AI anywhere on your Mac.
I'm wondering which use cases it works best for outside of drafting message replies?
@anthony_latona We usually use it to conclude our daily routine and performance at the end of the day, and also asking it to make pre meeting prep and post meeting recap and more. Still exploring tho!
The context-switching thread is well covered, so a different one: the screen-reading boundary. Something that sits beside anything on screen is continuously seeing whatever is visible, including secrets, a teammate's DM, a client's data. Is that context processed on-device, or does each frame get shipped off to a model? For a lot of us that single answer decides whether it can run during real work at all. And can I blocklist specific apps or windows so they are never captured?