AI agent handoff
What happens when one AI agent @mentions another in a project room: it answers at once, waits for a person, or is not passed on.
The words that turn up when you start calling a LinkedIn or messaging API and letting the work run without you — and what each one actually means once you are the one handling it.
What happens when one AI agent @mentions another in a project room: it answers at once, waits for a person, or is not passed on.
The lk_ string that decides which organisation you are, and therefore which LinkedIn accounts you may name.
The opaque handle for one connected LinkedIn account — and the one id in this API that is a security boundary.
The specialist AI agents every project room starts with: what each knows, when it answers, and what it cannot do.
LinkedIn pausing a sign-in to ask a question. The five kinds, and which ones you can answer from code.
The one bot that reads and posts in your Discord servers for you: added through Discord's own page, held to your servers, and never pinging anybody.
A campaign that keeps working after you stop calling: what the three kinds do, whose AI key writes the words, and why stalled is not paused.
A shared chat where your team, invited guests and AI agents work in one conversation, on your organization's own AI key.
Doing the same thing twice and getting one result — the property that makes retries and duplicate webhooks harmless.
The checkpoint with no code: LinkedIn asks for a tap in its mobile app, and nothing comes back to the request that asked.
A WhatsApp number connected the way WhatsApp Web is, with a QR code: no Meta app, no approval, and every message sent as the number itself.
The protocol that lets an AI assistant call this API directly, with one address and nothing to install.
Where a customer writes to you — WhatsApp, Instagram, Facebook, TikTok, Telegram, LINE, email, SMS, a website chat, X or Discord — and which API reaches each.
Software that runs on your own server or in another AI tool and joins one project room through the API with its own token.
Spacing calls out on purpose so an account looks like a person using it, and what that buys you.
A shared chat where your team, invited guests and AI agents work in one conversation, on your organization's own AI key.
Which address LinkedIn sees your sign-in coming from, why it matters, and when to bring your own.
Connecting an account with the cookie from a browser that is already signed in, and why the user agent has to match.
One inbox for every channel your customers write on, where a team answers, assigns, labels and resolves conversations — by hand, from code or with AI agents.
Our 0 to 100 measure of how much a LinkedIn account is actually read, the four parts it is made of, and what it is not.
Being told when something happens instead of asking over and over — and the two things everyone gets wrong about it.