a sourcing engine that SHOWS ITS WORK
RADR. turns a job description into a set of approved X-Ray searches, runs each one against Google, and indexes every public profile the engine returns. Nothing is scraped from behind a login, and nothing is hidden by a score. You read every query before it runs, and every result after.
- platforms searched
- 9A closed, reviewed list. Nothing is added quietly.
- approved query templates
- 16Fixed X-Ray patterns, filled from your constraints.
- search per query
- 1Each string gets its own Google call and its own receipt.
- profiles hidden by score
- 0Relevance orders the list. It never trims it.
- GitHub
- Stack Overflow
- Wellfound
- Behance
- Dribbble
- Resumes / CVs
- Multi-site pass
from job description TO SHORTLIST
One run takes a role from a document to a working pipeline. Four movements, each one visible and each one yours to edit.
- 01
Read the job description
Upload a PDF or DOCX, paste the text, or describe the role to the agent. Title, seniority, location, hard skills and domain terms are extracted. Soft skills are kept for scoring and never searched.
- 02
Build the queries
Approved X-Ray templates are filled from your constraints and your additional details. Every string is fitted to Google’s 32-word limit and shown to you, in full, before it runs.
- 03
Search and index
Each query is its own Google search with its own receipt: pages, credits, results. Every profile returned is indexed into the session. Relevance sorts the list, it never trims it.
- 04
Track and reach out
Shortlist, take notes, move people through a six-stage pipeline, and draft outreach in your company’s name from your own templates. Every action is logged against the candidate.
built for recruiters who READ THE QUERY
The engine is deliberately legible. Every rule below exists so that you can explain, to anyone, why a profile is on the list.
Approved templates, readable strings
Queries come from a fixed list of reviewed X-Ray templates. Read every string before it runs, edit it in the query editor, or build your own from syntax chips.
Every result, indexed
No liveness checks, no page scraping, no score cut-off. Only non-profile URLs and exact duplicates are dropped. What Google returned is what you see.
Your words carry the most weight
Additional details on the constraints form become required phrases and exclusions in every query. “Only fintech, no agencies” means exactly that.
Agent mode
Describe the role in conversation instead. The agent shows its reasoning, asks at most two blocking questions, and builds the same approved bundle.
Sessions that remember
Every run is a session: constraints, queries, receipts, candidates, notes and outreach. Re-runs refresh scores but never overwrite your edits.
Your key, your depth
Bring your own SerpAPI key; it is stored encrypted, server-side. Choose how many pages each query fetches, one credit per page, all shown on the receipt.
your first search IN SEVEN STEPS
From sign-in to a candidate in conversation. Nothing to install; a Google account and a SerpAPI key are all you need.
- 01Sign in and set up
- 02Start a session
- 03Confirm the constraints
- 04Read the queries
- 05Run the search
- 06Work the results
- 07Move people forward
- Step 01
Sign in with Google and set up in three fields
Your name, your company or organisation, and a SerpAPI key. The key is checked against your SerpAPI account before it is stored, encrypted, on the server. Outreach will later speak for the company you enter here.
- Step 02
Start a session from a document or a conversation
Upload a PDF or DOCX, paste the text, or open the agent and describe the role. Every session is saved with its constraints, queries, receipts and candidates, and appears in the history panel beside the sidebar.
- Step 03
Confirm what was extracted, then add your own details
Title, seniority, location, must-have skills and domain are read from the description; nothing is invented. The additional details box is yours: quoted or required phrases go into every query, exclusions come out of every query.
- Step 04
Read every query before it runs
The bundle is grouped by platform and shown as syntax tokens: site and filetype operators, phrases, OR groups, exclusions. Open any row to edit it, or build your own. Set how many pages each query should fetch.
- Step 05
Run it and follow the receipt
Each query is searched individually with live progress: pages fetched, credits used, profiles indexed. Runs continue in the background, so you can leave the page and come back to the results.
- Step 06
Work the results
Filter by platform or tier, sort by relevance, open a profile in the drawer, take notes and shortlist. Nothing is hidden by score: what you see is everything Google returned, minus duplicates and non-profile pages.
- Step 07
Move people through the pipeline and reach out
Six stages from New to Replied, on a board or in a table. Outreach is drafted in your company’s name from your own templates, with a connection note for the first touch, and every message is logged against the candidate.
