Every stage is grounded in structured facts about the lead. No imagination. No invention.
Generate a candidate pool from market + channel + vertical.
Structured extraction of channels, ecommerce maturity, traffic, content recency.
LinkedIn, X, Facebook follower counts and recency. Named leaders surfaced.
Five-component score with per-ICP weights. Hard disqualifiers tagged.
Twenty-eight-plus signals from a controlled, editable taxonomy.
Claude writes opener, pain angles, named leaders — only from facts.
Human-clicked enrichment. Enrichment credits only spend on leads worth it.
Brand-aware draft from Claude, with audience-matched blog credibility links.
Keyboard-first triage. J/E/Enter/X. 200 drafts in an hour.
Dual-brand SMTP. Your domains, your reputation, your throttles.
IMAP detects replies. Manual transitions through booked → won/lost.
Outcomes link back to versioned prompts. Next week's pitches are better than this week's.
how it worksFive-component scoring against your ICP rules — not a five-checkbox filter. Digital maturity, channel dependency, product fit, company size, accessibility. Weights editable per ICP. Hard disqualifiers tagged before any model spends a token on this lead.
read the deep page
Twenty-eight-plus signals in a controlled, editable taxonomy. multi_system_scatter, manual_po_entry, marketplace_dependency. Derived from web audit + LLM + manual override. Pitch angles can only fire from observed signals.
No generic merge fields. Solution templates triggered by pain-signal intersection — pitch angles grounded in the specific pain observed on this lead's site. If the input is missing, the model returns empty rather than inventing.
read the deep pageReview hundreds of unique, pain-grounded drafts in an hour. Keyboard-first triage. J next, E edit, ⏎ approve, X skip. Your domains. Your reputation. Your call on every Send.
Every LLM call writes prompt id, input presence, missing-required, tokens, engine, model, output meta. Funnel events on the lead link back to the exact prompt variant that drafted it. After a quarter, you can answer "which pitch_brief variant has the highest promote-to-reply rate on devtools × kubernetes" from SQL — not guesswork.
how the loop worksprompt_id: pitch_brief_v14 scope: devtools · kubernetes · b2b input_presence: {site:1, social:1, channels:1, kb:1} missing_required: [] engine: anthropic model: claude-sonnet-4-6 tokens_in: 4,127 tokens_out: 890 output_meta: {templates_fired: [t_07, t_11]} ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ downstream: email_drafts.pitch_prompt_id → pitch_brief_v14 funnel_replied → true (+5d) funnel_call_booked → true (+9d)
Six things you'd build anyway, included.
You bring the domains. We send and detect replies on both. Per-brand signatures, From-chains.
In-page prospecting. Domain match, cross-page crawl, pitch-brief — writing into the same lead DB.
Personalized cover-slide PDFs auto-generated with their logo + your pitch snippet. Tracked share links.
Site traffic ingest. AI-generated weekly digest tying outreach activity to inbound signal.
RRULE recurrence. FullCalendar view. Plan-Week kanban for content + follow-ups.
Admin / rep roles. Per-rep allowed-lead-type RBAC. TOTP 2FA with backup codes. Lead assignees.
Drop a CSV. Pick a vertical pack. Watch eleven stages run live.