> ## Documentation Index
> Fetch the complete documentation index at: https://docs.revcenter.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# New Search

> From a plain-language ask or a JD file to a ranked, qualified candidate list

New search (`/intake`) turns a role description into a validated candidate search.

## Input

Describe the role in your own words, paste a job description, or attach a **PDF, DOCX,
TXT, or MD** file with the paperclip. As you type, five chips confirm what the system
recognized — Location, Job Title, Years of Experience, Industry, Skills. Location accepts
a city ("Detroit, MI"), a state, "USA" for nationwide, or "remote". Written-out numbers
("Eight (8) or more years") are understood.

Two modes: **Candidates** (find people for a role) and **Clients** (find firms with open
roles — calibrates at the firm level, and pauses for your approval before the paid
org-chart scan). The model selector picks extraction strength: Haiku (fastest), Sonnet,
or Opus.

## What runs when you send

<Steps>
  <Step title="Pool search — free">
    Your workspace's already-enriched candidates are searched semantically first. Warm
    matches cost nothing and arrive already enriched; matches that fail this search's
    gates are skipped, not imported.
  </Step>

  <Step title="Discovery">
    Google X-ray queries built from the criteria. Small markets automatically widen:
    the commute radius scales with market density, and searches near small towns add
    the nearest sizable city ("Findlay" also searches "Toledo").
  </Step>

  <Step title="Enrichment">
    HarvestAPI scrapes each new profile (cached 60 days — re-surfaced profiles are
    free). Anonymized "LinkedIn Member" results are dropped before costing anything.
  </Step>

  <Step title="Gates, scoring, and the qualification loop">
    Hard gates reject clear mismatches with a named reason; everyone else gets a
    JD-weighted score. The pipeline searches deeper until it holds at least 10
    enriched candidates at 85+ (or your requested count), then reports either the
    economics line or a scarcity report naming what's binding. See
    [Search Quality & Scoring](/guides/search-quality).
  </Step>
</Steps>

## Reading the results

* **Tiers**: A (85+, qualified) · B (70–84) · C (55–69) · D (below, or gated).
* **Gated rows** show their reject reason inline ("out of area", "missing required
  skill: cpa") — a zero always explains itself.
* **Signal chips**: tenure, move-ready, people leader, open-to-work, and per-search
  fit signals.
* Click any row for the full profile — and record an **Accept / Maybe / Reject**
  verdict; these labels calibrate scoring to your team's standards over time.
* **Find more** searches deeper on the same scan; **CSV** exports the ranked list with
  subscores, tier, and reject codes.