THINK OUT LOUD
A little conversation.
A new possibility.
Ask, explore, or think out loud. Connections can follow when they are useful.
Voice uses your browser’s speech service. Dictation is reviewed before sending. Sharing with people requires separate approval.
CLEAR INTENT, MORE CONTROL
Your possibilities.
Approve an intent, then ask your agent to send private introduction requests.
HUMANS HAVE THE LAST WORD
An introduction.
At your pace.
Review the exact disclosure before accepting. Private histories are never attached.
GROW YOUR CIRCLE
A familiar name.
A new connection.
Add someone you know.
Only exact known usernames work. There is no public user directory. Both people must accept before friend chat opens.
MORE CAPACITY. THE SAME PERMISSION.
Choose your membership.
Monthly AI credits and new connections. Every introduction still needs a human’s approval.
Your delivery preference controls where introductions can happen. Membership does not limit distance. Voice requires a configured provider; no plan promises unlimited usage. Billing must be configured before checkout is available.
YOUR SPACE, YOUR CONTROL
Privacy is a feature.
MEMBERSHIP & USAGE
Your monthly allowance
One successful text AI reply uses one credit, including memory processing. Local fallback replies and failed provider calls use none. Catalogue answers use none. A new connection counts for the initiator when both people accept, or when a business owner approves the first follow-up reply. The same pair is counted once across introductions, friendships and business follow-ups. Pending requests and questions, incoming connections and existing human chats use none.
YALVI SAFETY
You can still sign out, export or delete your account, and manage an existing subscription.
What your agent can share
Only your approved intent excerpt and display name, when you request an introduction. Another person’s excerpt becomes available to you only after they accept.
Private chat and precise coordinates are excluded from every introduction. With AI enabled, OpenAI processes chat messages for replies and safety checks. TypeSafe’s Jev reviews selected messages, opted-in memories and shared-room messages for useful context and safety, and approved excerpts for matching. Matching never uses another person’s private chat history. The platform server can decrypt stored data. This is not end-to-end encryption.
Deleting an account removes its chat, intents and connections. Cancel billing first. Pilot reports are recorded but require an operator review process before public launch.