Customers

Lead Discovery

Lead databases sell you job titles. What you actually want is intent: the person who posted yesterday asking for exactly what you build. That signal lives scattered across X, Reddit, Hacker News, Xiaohongshu, LinkedIn threads, and comment sections — too many places to watch by hand. Discovery watches them for you, with a written definition of who you're looking for.

The customer profile brief

Everything starts from the brief — an editable document the discovery agent works from. Generate brief drafts it from your product assets; you refine it. It contains:

Target personas — who buys, in plain language.
What counts as a valid lead — one rule per line. ("Asked for a tool recommendation", "complained about [competitor] pricing".)
Never surface — the exclusion list. (Agencies, students, job seekers.)
Search plans — the concrete queries and angles the agent will pursue.
Hunting plan — where and how, per channel, with a tactic each: intent search, ask threads, hot comments, competitor watch, or web sweep.

The brief is a living document. Regenerate from product assets rebuilds it after your positioning changes, and your manual edits survive regeneration. Ten minutes tightening the "valid lead" and "never surface" lines is the highest-leverage act in the whole outbound engine — every run downstream inherits its precision.

Seeds

Seeds are starting points you already have: paste handles, profile URLs, or emails (one per line), or paste a post URL to mine — the author and everyone in the comment section become candidates. Importing seeds costs nothing; seeds are inputs, not leads.

Running discovery

Press Start now. Each run picks one focused direction from the hunting plan — one channel, one tactic — rather than spraying everywhere. The hunting plan keeps a yield ledger per channel ("6 valid over 4 focus runs"), and future runs favor the channels that actually produce for your product. Candidates found are screened against the brief, and each surviving person is verified as a real individual before being delivered.

You can start 3 manual runs per day; counters show today's usage and total found.
Ongoing discovery requires a plan; free brands get the taste allowance to evaluate quality first.
Runs continue in the background — the Activity feed shows progress, and a notification fires when the run lands.

Reading the results

Every lead row carries:

Match score and a signal type: actively seeking, competitor complaint, pain point, or persona fit. The first two are the hottest — those people are asking to be contacted.
The match reason with quoted evidence, linked to the source post. You can always click through and read the person's actual words before deciding they're worth your time.
Identity and links — role, what they build, follower count, and profile links, with a tag showing where the name was resolved (X, GitHub, LinkedIn, their site, seed, and so on).
Copy suggested opener — a first message grounded in their specific evidence, ready to adapt and send.

Filters: search, role (founder / indie / investor / maker), location, minimum followers, and has email / has X chips.

Per-row actions:

Shortlist (★) — pin the promising ones; the shortlist is your working set for outreach.
Deep dive (⛏) — commission a full dossier on this person. See Deep Dive Dossiers.
Unlock email — contact details unlock per lead (padlock icon when locked).

What you pay for

Billing is per delivered, verified person — 25 credits, adjusted by a market multiplier (from 0.5× for high-volume B2C audiences to 2× for scarce B2B ones). Searching, screening, and every candidate that failed verification cost you nothing. See Credits & Metering.

Getting better leads

Tighten "What counts as a valid lead" after every batch you review. It's the single input the judge reads most literally.
Mine your best community thread as a seed post — comment sections of "what tool do you use for X" threads are dense with intent.
Watch the yield ledger and prune channels that stay dry; the focus system will follow your lead.