Settings has every integration this app can use — Gemini/Claude for AI, Outscraper/Apify for scraping, Hunter/MillionVerifier for email finding, Instantly for sending, GoHighLevel/SignalHouse for calls+SMS. Nothing is required to start — most tabs work with zero keys, they just show honest 'not configured' states instead of guessing. Add keys as you need them, then run [Test all connections] on Settings to confirm each one actually works.

Prospecting is where a run starts — pick a niche + city (or a saved preset), and it scrapes real local businesses, audits each one's site for missing tracking pixels or a broken/absent site, and scores them. Results shows every lead from every run with the full audit breakdown, honest email-finding status, and one-click actions (find email, call, push to a sequence).

Pipeline is the lifecycle view — new → contacted → replied → meeting → won/dnc. Sequences sends the actual outreach emails (Flow Alpha/Beta/Gamma, matched to what's wrong with the lead's site). Inbox shows real replies + a unified thread per person across email/SMS/calls. Calls is the dial mode with the Sales Playbook, Live Call Copilot, and Call Scorecard grading.

Once a lead is Won, Ad Audit Pro's Workbench + Clients tabs manage the ongoing ad account — real Meta data, explainable flags, a recommendation queue with an apply/rollback trail. AI Department offers AI-product upsells to healthy clients after 90 days. Sales HQ is the day-to-day cockpit: Money Scoreboard, Manager Digest, Power Hour, goal streaks.

The cockpit — today's activity, Money Scoreboard (dials → $ pipeline), skill progression from graded calls, Power Hour, daily goal ring and streaks.

Run a new scrape: pick niche(s) + market(s), see a live progress bar, an exclusion preview, and a cost estimate before you commit an API call.

Every lead from every run — audit grade, email-finding status, bulk actions (find emails, push to Pipeline, verify), an honest one-line brief per lead.

The 9-stage lifecycle board (new → dnc) with channel-lane filters, archive/delete, and a Send-Next-25 preview before a batch actually goes out.

Real replies synced from Instantly, plus a Unified threads view merging email + outbound SMS + calls into one timeline per person (inbound SMS is honestly absent — no webhook connected).

Dial mode: Sales Playbook, AI Pre-Call Brief, Live Call Copilot for objections, Call Scorecard grading (also extracts real commitments made on the call), Voicemail Drop templates.

The Night Shift's morning report — what ran overnight (audits, email finds, ad rescans, Platform Watch) plus the Manager Digest (today's activity, a coaching tip on the weakest graded call).

The 3 outreach flows (Alpha/Beta/Gamma, matched to what's broken on the lead's site) with health cards, guided send, and a test-send-to-yourself button.

Ad Library search for any advertiser/niche — winner badges, Claude hook analysis, a swipe file of good ad copy, and a watchlist that flags when a competitor's ads stop running.

The ads-operations suite for won clients — Workbench (real Meta account data, explainable flags, Creative Doctor, a 6-step 'clicks but no calls' diagnostic), Audit Anything, Library, Clients, Queue (recommendation approve/apply/rollback), Negative Keywords, Search Term Engine, Benchmarks.

The upsell funnel for AI products to healthy Won clients (90+ days) — a 6-stage funnel from Won to Live, gated on real client-health data, never a fabricated 'ready' status.

Every real promise Claude found in a call scorecard transcript — bucketed Now/Today/Digest/Weekly by urgency, so nothing said on a call gets forgotten.

Real Google/Meta ads-industry news, classified by Claude (breaks something / costs money / new capability / noise), plus a countdown board for known migration deadlines.

The library of scrapeable business types — search/sort, per-niche stats, and a Claude-powered 'forge' that can build a new niche definition from just a name.

Every API key and config value, grouped by purpose, with a [Test all connections] health check and per-key status lights.

A handful of the app's automatic overnight behaviors (finding emails, checking for ad changes, sending reminders) can each be set to Observe (skip and just report), Propose (compute it but hold it for a one-click review on the Staff Desk page), or Act (today's default — fully automatic, no approval needed). Set per-desk in Settings under 'The Staff'.

Anything left in Propose mode shows up on the Staff Desk page — review and apply or dismiss each one individually.