How to use Jev in n8n to sort your LinkedIn inbox

What is Jev?

n8n's post calls Jev the first System One model from TypeSafe AI n8n blog: Build a morning AI news brief with You.com, Jev, and n8n. It is not a chat model. In the post's words n8n blog: Build a morning AI news brief with You.com, Jev, and n8n:

"You hand it a state and a set of typed questions, and it returns typed values with calibrated confidence, in well under a second."

n8n blog: Build a morning AI news brief with You.com, Jev, and n8n

The same post says Jev generates no text, and that it was in early access when the post was written, with a TypeSafe API key needed to use it n8n blog: Build a morning AI news brief with You.com, Jev, and n8n. In the post, Jev scores news articles. This page is about pointing the same idea at your LinkedIn.

Why a sorter suits a LinkedIn inbox

Much of the work in a LinkedIn inbox is deciding what deserves an answer: a sales pitch, a recruiter, a real lead, an old friend. That call is made once for every message and every invitation.

n8n's post makes the same point about news: deciding what is worth reading is the expensive part, because it takes a judgment call on every item n8n blog: Build a morning AI news brief with You.com, Jev, and n8n. A model that only answers questions can make that first cut. You then read what is left.

What the workflow looks like

1. A Schedule Trigger runs it each weekday morning.
2. An HTTP Request node asks your LinkedIn who is waiting on a reply from you, or which invitations you have not answered yet.
3. For each one, another HTTP Request node reads the conversation or the person's profile.
4. Jev gets that as its state, with your questions.
5. An If node keeps what passes, and Slack or email posts the short list to you.

• who_is_waiting_on_meList the people whose last message you have not answered.
• get_conversationRead your whole conversation with one person.
• list_received_invitationsList the invitations you received and have not answered.
• get_profileRead one person's profile in short, with their recent posts and their company.

All four only read. Nothing goes out on LinkedIn from this workflow.

Which questions to ask Jev about each message

n8n's post asks three separate questions about every article rather than one combined verdict, so each can be tuned on its own n8n blog: Build a morning AI news brief with You.com, Jev, and n8n. The same shape works here. For a message, for example:

• Is this person asking me something, or only pitching?
• Is it about the work I do?
• How much does it matter to me this week, given a sentence about what I sell or what I am hiring for?

For an invitation, ask whether the person's title and company fit who you want to meet. Keep the threshold in one place, so you can raise it after a week of real mornings.

Who writes the reply?

Not Jev, which writes nothing. You answer the short list yourself, from LinkedIn, or from your AI app with the same LinkedIn connected:

• "Read my conversation with the first person on this list and draft a reply. Show it to me before you send it."
• "Accept the invitations from the three people I marked, and decline nothing."

• send_messageSend one message to a 1st-degree connection, or into an existing conversation, now.
• respond_to_invitationAccept or decline one invitation you received.

If you would rather not build any of this, the LinkedIn AI agent sorts who to talk to next and asks you before anything is sent. On the free plan LinkedIn disconnects after a while unused, so a morning workflow can find it gone; the plan that keeps it connected is on the pricing.

Where to set it up

The key, the address and what to type in each HTTP Request field are on the n8n page. Each action, with its daily limit, is on the actions page.