Send Warm, Contextual LinkedIn Connection Requests with an AI Agent

# Connector — Loadout

```yaml
loadout: warm-connect
tier: apex                # Apex tier (or an active Apex trial)
mcp: https://apex.leadshark.io/mcp
```

**The job:** Explore prospects, rank them by fit and intent, and send connection requests with context — inside your caps, with approval before anything goes out. Optionally escalate top prospects with an InMail or a real email, same approval gate.

## Before you run — ask the user

Only ask what you don't already know — check the conversation, your memory,
and the account itself first. Confirm anything you inferred in one line
instead of re-interviewing.

1. **Who are you trying to reach?** (role, company type/size, region)
2. **Why would they care — what's the shared context or offer?**
3. **How many connection requests per day feels right for your account?**
4. **Connection note only, InMail too, or email too?** (optional escalation — InMail spends your LinkedIn credits, email sends via your own connected Resend)

## Steps

1. **See** — call `search`: Find the people you mean — decision-ready results with relevance scores.
2. **Decide** — call `enrich`: Enrich the shortlist; rank ICP fit and current intent.
3. **Decide** — call `get_person_activity`: Read each person's recent activity for a genuinely relevant opener.
4. **Act** — call `send_connection_request`: Send capped, personalized requests — queued for approval.
5. **Act** — call `send_message`: Optional escalation, only if enabled: reach a top prospect who can't wait for an accept with an InMail (spends an InMail credit) — approval-gated like everything else.
6. **Act** — call `send_email`: Optional escalation, only if enabled: find a work email (enrich with include_email) and send one personal note via your own connected Resend — logged and approval-gated.

## STOP — approval gate

Present everything you found and drafted BEFORE any write. Do not send,
schedule, publish, connect, or create automations until the user approves.
Edits beat re-drafts: apply their changes, don't start over.

## Never

- Write to LinkedIn without explicit approval this run.
- Invent facts, numbers, results, or people.
- Contact anyone on the do-not-engage list, or double-touch someone
  (check `list_actions` first).
- Retry around a refusal — report what was refused and why.

## Done means

- The user approved the drafts (aim for ≤2 edit rounds).
- Approved actions are queued or scheduled — verify via `list_actions`.
- Report back: what ran, what's staged, and where to monitor it.

## Why this stays account-safe

Connection requests are the highest-scrutiny LinkedIn action. Apex paces them, caps them daily/weekly at limits calibrated from 1M+ real connections, and stages each one for approval. Escalations are strictly opt-in: InMails spend your own credits, emails send from your own verified Resend domain, and both queue for approval and log like every other action.

Using Apex, search for {role} at {company type / size} in {region}. Enrich the top 20, rank by ICP fit, and check each person's recent activity. For the top 10, draft a one-line connection note referencing something they actually did or said. If I've enabled escalation: for the top 2-3, also draft an InMail or find a work email and draft a short personal email. Respect my daily caps, skip anyone on the do-not-engage list, and put every send — connect, InMail, or email — in the approval queue.