---
title: Import contacts into Venom People
description: Bring in people from files you already have, Google Contacts, iCloud and Google vCards, LinkedIn connections, WhatsApp and Telegram exports, HubSpot and Pipedrive, with a review that merges duplicates and history that seeds relationship strength.
section: cloud
order: 164
audience: [user, operator, agent]
sources: ["services/nodejs-cloud/app/(people)/people/import", services/nodejs-cloud/components/people/Import.tsx, services/nodejs-cloud/components/people/import-reader.worker.ts, services/nodejs-cloud/lib/people/import, services/nodejs-cloud/test/fixtures/people-import, services/nodejs-cloud/migrations/0176_people_graph.sql, services/nodejs-cloud/migrations/0167_people_network.sql, docs/adr/0057-venom-people.md]
features: [cloud.page-people]
---

Open **People → Import** (`/people/import`). Venom reads files **you** export from the tools you already use. It
fetches nothing from LinkedIn, WhatsApp, Telegram, Google or anyone else, and it does not scrape. The file is read in
your browser, in a background worker; only the people it describes go to Venom, in batches, after you review them.
Everyone is imported **private to you**; share them later from [the network](/docs/cloud/people-network.md).

## What you can import

| Source | Get the file | What Venom reads |
|---|---|---|
| Google Contacts | contacts.google.com → Export → Google CSV | name, every email and phone, company, title, notes, labels (as tags) |
| iCloud or Google vCard | iCloud: select all → Export vCard; Google: Export → vCard; or a phone's "share all contacts" | any number of cards, vCard 2.1, 3.0 and 4.0 |
| LinkedIn data export | LinkedIn → Settings → Data privacy → Get a copy of your data → the whole `.zip` (or `Connections.csv` and any of the other files) | see [LinkedIn](#linkedin) |
| WhatsApp chat | in a chat: Export chat → Without media → the `.txt` or `.zip` | who wrote and when: every active month becomes a touch, with who reached out |
| Telegram | Telegram Desktop → Settings → Advanced → Export Telegram data → JSON → `result.json` | personal chats the same way, and the phone numbers in your contact list |
| HubSpot | Contacts → Export → CSV | name, email, phones, company, job title, city; the last activity (else create) date becomes a touch |
| Pipedrive | People → Export → CSV | name, organisation, email, phone, job title; the last activity (else created) date becomes a touch |

Any other CSV with a name column also works. A phone number without a `+` can be given a country code on the first
screen (for example `65`); without one it is kept as typed.

## LinkedIn: the whole export {#linkedin}

Pick the `.zip` LinkedIn emails you, or its CSV files one by one (select several at once). The zip is opened in your
browser; nothing is sent to LinkedIn or anyone else, and files Venom does not use are never read. It reads:

| File | What Venom does with it |
|---|---|
| `Connections.csv` | the people: name, profile URL, email, company, position. The few lines of notes above the header are skipped. **Connected On** is saved as the date you connected and as a `met` touch |
| `Positions.csv` | **your own** employers, to show where you and a connection worked at the same company. Not added as people |
| `messages.csv` | for each person you wrote with one to one, **one touch per active month** (the last two years) saying how many messages and who wrote. Groups are left out. **Message text is not read** unless you tick *LinkedIn messages: keep a few short lines per month* |
| `Invitations.csv` | who invited whom to connect, and when |
| `Recommendations_Given.csv`, `Recommendations_Received.csv` | a small, strong touch on the person, and the tag `you-recommended` or `recommended-you` |
| `Endorsement_Given_Info.csv`, `Endorsement_Received_Info.csv` | how many skills each person endorsed or you endorsed of theirs, as one touch |

Company Follows and everything else are ignored. Messages, invitations, recommendations and endorsements attach to the
people in Connections.csv, matched by profile URL then by name; a name that is not a connection is counted and left out.
These touches feed [the strength score](/docs/cloud/people-keep-warm.md), so people you really talk to rank above ones you only
connected with.

People you already have are matched by LinkedIn URL first (in any spelling: `www`, a country subdomain, a trailing
slash), then by email, then by name and company. Before anything is saved the review shows four counts: **New**,
**Updated** (the file adds something to someone you know), **Unchanged** (nothing new: a re-import skips them) and
**Skipped**. Importing the same export again changes nothing. Everyone shows up in [the graph](/docs/cloud/people-graph.md).

## Chats: counts and dates, not messages

A chat tells Venom how warm a relationship really is. For a chat with one person Venom stores **one touch per active
month** (the last two years) saying how many messages and who wrote, so recency, frequency and reciprocity in
[the strength score](/docs/cloud/people-keep-warm.md) start from real history. **Message text is not imported** unless you tick
*Keep a few short message lines per month*. In a group, only the people you pick are added, with one low-weight note
each; a group is a weak signal. WhatsApp's export does not say which side is you: pick it in the list (the file name
usually settles it). Dates such as `01/02/2024` are read day first when nothing else in the file decides.

## Review, then import

Before anything is saved Venom shows who you already know. A shared email, phone, Telegram handle or LinkedIn URL is a
**strong** match; the same name at the same company is a **likely** one. For each row choose **Merge into …**, **Create
new** or **Skip**, or use *Merge all matches*, *Skip everyone I know* and *Create all as new*. The default is to merge
a match and create the rest. A merge fills empty fields, adds handles and tags, and never overwrites what you wrote.

Importing the same file twice adds nothing: each imported touch carries a key, so a second run merges by handle and skips
touches already there. Imports spend **no AI**: no context cards are refreshed per row, and the new people are embedded
afterwards in the background with your organisation's embedding model (if it has one). Strength is recomputed for
everyone touched, so imported history shows on Today straight away.

## Agents

`people_import_run` takes a file's text (CSV, vCard (LinkedIn: `Connections.csv` only; use the app for the whole export), a WhatsApp `_chat.txt` or Telegram's `result.json`), detects the
format or takes `source`, and without `confirm` only reads: how many people, which you already know, nothing saved. With
`confirm` it saves: strong matches merge, likely ones follow `on_likely` (`merge` by default, or `create`, `skip`), the
rest are created. Over MCP and REST (`POST /api/v1/people/import`) a call takes at most 500 people; use the app for
bigger files. See [the workspace MCP server](/docs/cloud/people-mcp.md).
