Prospecting & the agent

Prospecting finds companies that fit a venture, researches them, and prepares outreach — inside the law, with a person approving everything that goes out. It lives in Lead Engine → Prospecting (sidebar: Customers & pipeline → Lead Engine, then the Prospecting tab).

Where to start

Prospecting opens on Agent. The usual order for a new venture:

1. Senders: choose the mailbox each venture's email goes out from (only the mailbox's owner can choose it), and its LinkedIn profile.
2. Targeting: what the venture sells and who it's for, so companies are scored for that venture.
3. Agent → Find new companies, then Research companies, then Prepare outreach.
4. Outreach: read each draft and approve it. Nothing is sent without that.

The agent: missions

The agent does fixed jobs — missions — for one venture at a time. Pick the venture, set the mission's options, and press Run now, or turn on Every day at and choose the hour (in the venture's time zone). Each card shows how much there is to work on and what a run will cost before it starts; the history below lists every run, what it found and what it was charged.

• Find new companies: adds companies that aren't in your pool yet — from your LinkedIn network (import your connections in My network), from web search shaped by the venture's targeting (add one keyword such as "clinics" to narrow it), and, if you tick it, from companies that wrote to your connected mailboxes (those companies become visible to your team). A company is added only if its website answers.
• Research companies: reads each company's own website and job board, keeps what it found with the words and the link, finds the contacts the company publishes, and scores the fit with every reason shown.
• Prepare outreach: drafts a message to the best person at each of your top-scoring companies, under the consent rules. The drafts wait in Outreach for you.

Replies are read and sorted automatically every 15 minutes.

Credits and the monthly allowance

A unit is work delivered: a company added, a company researched, a draft made. Units come first from your plan's monthly research allowance, then one credit each (buy credits on Billing). A unit that fails, or that the consent rules don't allow, costs nothing. A run never asks for more than you can pay for, and stops when credits run out.

Outreach and approval

Every message is a draft until a person approves it. Email goes out from the venture's mailbox, inside its sending hours and daily cap; a new mailbox warms up (a few a day, rising daily). LinkedIn messages are copied and sent by you — LinkedIn doesn't allow automated sending.

In Israel, the first email to someone who hasn't agreed to hear from you may only ask whether they want what you offer. A reply is never taken as consent by itself: the mailbox's owner confirms it.

Follow-ups after a yes

Follow-ups is a short sequence (up to 5 steps, days apart) the mailbox's owner writes and approves once. It starts only for people whose yes the venture holds, and stops by itself on any reply, a request to stop, or a bounced email.

Do not contact and consent

Do not contact lists every address, domain, phone and profile nobody may contact. A "remove" reply, the unsubscribe link in every email, a bounce or a complaint lands there automatically. By default a yes belongs to the venture it was given to and a no stops every venture; an owner or admin can change that on the same page.

Mailbox health

Senders shows each mailbox's warm-up day, today's cap and its last 7 days. Too many bounces or requests to stop pause the mailbox until its owner resumes it, so one bad list can't harm your sending reputation.

Your own AI agent (MCP)

An outside agent (Claude and others) can work the same pool through MCP with an API key that has the Prospecting agent (MCP) scope (Settings → API keys). It can find, research, add and propose — it can never send, and it can never record a yes.