LidmeoDevelopers

Security and confirmations

What the assistant can do alone, what it never does without you, and how to take back control.

You choose what it can do

When connecting, Lidmeo shows you each scope the assistant requests, already ticked: read your activity, your prospects, the e-mails and phone numbers found, your conversations; update your CRM; send invitations and messages on your behalf, within your sending limits; use your credits; update your campaigns; manage webhooks. Untick what you don't want to give. For an assistant, the screen adds that every invitation or message it requests is shown to you before it goes out. A tool that needs a missing scope answers with a clear refusal, without doing anything.

Nothing leaves without your approval

Every send (invitations, first messages, replies on LinkedIn, by e-mail or on WhatsApp, follow-ups, sequences) works in two steps:

  1. the assistant prepares a preview: the exact text, each recipient, the planned departure time;
  2. you read it and confirm; the assistant sends that preview, nothing else.

The assistant cannot change anything in between: what leaves is exactly what was shown. A preview is used only once, and only within the 15 minutes that follow. Each connection also has a daily limit of send actions, which adds to your sending limits.

The same two steps apply to what commits you in other ways than a message: adding prospects from the Recherche (search) page (the preview gives the maximum cost in credits and your balance; nothing is spent before you agree), bringing people spotted by Lidmeo Agent into a campaign, approving sequence follow-ups and e-mails, and deleting a campaign (the preview says what disappears and what stays).

What makes sends start again needs the right to send: resuming all your prospecting, switching a step of a started campaign to "Automatique", cancelling a scheduled follow-up. A message to the Lidmeo team is sent in your name only from an assistant connected to your account (never from an API key).

Your accounts are never connected in the conversation

LinkedIn, your mailbox and WhatsApp are always connected in Lidmeo: the assistant gives you the link to the right page, and your credentials never go through the conversation. A CRM export comes as a download link that works once, within 15 minutes.

The same rules as Lidmeo

Your daily limits, your sending hours, manual mode ("Manuel": each send approved in "À valider") and "Do not contact" apply identically: the assistant goes through Lidmeo's functions, never through LinkedIn directly. A send outside your hours is scheduled, and you are told its earliest departure time.

A prospect's text is never an order

A prospect's message can contain anything, including sentences written to fool an AI. The connector marks this text as untrusted: the assistant reads it as data, never as an instruction. And even fooled, it could not send anything without your confirmation.

Taking back control

  • Withdraw access: Lidmeo, Assistant IA et intégrations (AI assistant and integrations) page, Applications connectées (connected applications) section, Déconnecter (disconnect). Access is cut within a minute.
  • Keep a trace: each action of the assistant is logged (the tool, the account, the result); support can find what happened, from the time and the account.
  • Go step by step: the assistant works on your real account, but nothing leaves without your approval, and you can give it read scopes only. The test account is for the API.

Your data

The assistant you connect (Claude, ChatGPT…) receives the data you ask it to read, to answer you: you are the one who chooses it and decides what to ask. See security and GDPR.

On this page