Run Managed LinkedIn Lead Pipeline

Automates the end-to-end lead pipeline (search to human judgment to storage to scheduled outreach) so invites scale across multiple accounts without manual intervention. Your agent qualifies leads against your ICP in Phase A, then Phase B runs on a background schedule sending invites at your pace and withdrawing stale pending requests, all while persisting state locally so interruptions do not lose progress.

After first-run setup (register accounts, set per-account limits, enable the scheduler), the workflow is two phases. Phase A (import): you give your agent a LinkedIn or Sales Navigator search URL; your agent prepares candidates, interviews you for an ICP (if not already set), qualifies each candidate via sub-agent judgment, and commits qualified leads to the local database with round-robin owner assignment across active accounts. Phase B (background): the OS-level scheduler wakes periodically; for each active account within its active window, it sends one invite (if the daily limit is not reached and the min interval has passed), checks pending requests older than the max-days threshold, and withdraws stale ones or marks them connected. All state persists in SQLite. Output: a queryable local lead database with status tracking, run history, and full auditability per lead.

Requires Node.js ≥ 20, @linkedapi/linkedin-cli, and npm dependencies (better-sqlite3, a native module requiring build tools). The background scheduler uses OS-native mechanisms (launchd on macOS, systemd-user or cron on Linux, schtasks on Windows) and wakes every 5 minutes (internal; user-facing behavior is controlled by per-account limits). All state is local; no external services beyond linkedin-cli. Timestamps in SQLite are UTC; invite daily limits reset at local midnight. Sales Navigator search support requires an active subscription.