Skip to content
ZoNuLyDOCS
ZoNuLy · Documentation

Five thousand companies. One person to ask.

Figures of 23 Oct 2026 · reading live…
  1. 8,085roles read
    from company-owned boards and careers pages
  2. 5,910companies tracked
    graded on funding, pay power and hiring
  3. 574complete
    hiring proven, and someone there to ask
  4. 25asks a day
    spread through the working day
  5. 1person, one email
    sent only after you approve it

ZoNuLy finds engineering roles at underrated, recently funded companies, proves they are really hiring, finds the real people who could refer you, and writes each of them one honest email — then stops and waits for you. You spend ten minutes a day approving; it spends the other nineteen hours doing the rest.

It runs on one laptop: a Python pipeline over a single SQLite file, a FastAPI server, a Next.js dashboard, and a handful of background jobs that keep filling the gaps around the clock. Models are reached through OpenRouter behind a cost ledger; nothing runs a local model. Every fact it shows sits next to the sentence it came from, and anything it could not prove is shown as missing — never guessed.

The pipeline in one picture#

Thirteen stages take a company nobody has heard of to a sent referral ask. Twelve are automated. One — deciding that this email goes to this person — is yours, and the system is built so that decision takes seconds because everything before it is already done.

  1. 01DiscoverCompanies from ATS boards, YC, Wellfound, funding news, India sources
  2. 02FillDescription, LinkedIn page, funding, pay power — each with its evidence
  3. 03Hiring proofAn open role on their own careers page or job board
  4. 04Roles & scoringRoles scraped, matched to the resume, ranked by selection odds
  5. 05DecideP(offer) per company; a shortlist per city and one company a day
  6. 06PeopleEngineers and founders who work there now, in the company's own city
  7. 07AddressesReal addresses first; a guessed one only when nobody ready has a real one
  8. 08DraftOne honest email per person, from evidence — or no email at all
  9. 09ReviewYou read, edit, approve or reject. Nothing sends without this.
  10. 10SendFrom your Gmail, 25 a day, at most 3 into one company, spread like a person
  11. 11RepliesRead back, classified, bounces caught, the brake applied
  12. 12Follow upExactly one, after five quiet days — back through review
  13. 13ScheduleA “yes” with a time becomes a calendar event, clashes flagged
Read left to right, top to bottom. The solid step is the only one a person performs.

What makes it different#

It looks where others don't#

Most good early-career roles at funded startups never reach the big job boards — they sit on the company's own Greenhouse, Lever or Ashby page. ZoNuLy reads those boards directly, and deliberately skips the household names whose roles get thousands of applicants. See Targeting.

It asks a person, not a portal#

A cold application is one of eight hundred; a referral from an engineer puts you in a stack of ten. So the product isn't a list of jobs — it is a company that is hiring, a role you fit, and a named person who works there now with an address that will reach them. The dashboard calls that “ready to ask”.

It decides before it spends#

A few thousand companies is too many to work on. The decision brain gives each Bengaluru, Hyderabad and remote company one number in plain words — say, “19% chance of an offer, about 1 in 5 companies like this” — keeps the best 100 / 50 / 100 on a shortlist, and hands you one company a day. Paid data is bought only for shortlisted companies that still lack it (the mixer), and what the shortlist keeps asking for that your resume lacks becomes a ranked learning plan (Upskill).

It refuses to make things up#

Every description, funding round, address and claim of employment carries its evidence. An email that would need a fact the system doesn't have is not written. An address is labelled verified, pattern-guessed or scraped, and the label is never dropped. See Guarantees & guardrails.

The one rule above all others
Nothing is ever sent without a human approving that exact email. No background job can send an unapproved message, and the daily cap holds even when you ask it to go faster.

Where to go next#

Start hereWhy ZoNuLy existsThe problem, the bet behind it, and the rules it never breaks.Start hereQuickstartInstall, run the API and the dashboard, and turn on the schedule.The pipelinePipeline at a glanceFrom an unknown company to a sent referral ask, stage by stage.The pipelineThe decision brainWhich companies to spend the days on: P(offer), the shortlists, one company a day.Using the dashboardDashboard tourEvery page in the app, what it shows, and what you can do there.ReferenceCLIEvery command in scripts/run.py.

How these docs are organised#

  • Start here — what it is, why it exists, and how to run it.
  • Concepts — the vocabulary, the targeting rules, the guarantees and the settled decisions.
  • Architecture — the stack, the data model, the background schedule and the knowledge graph.
  • The pipeline — one page per stage, from discovery to the calendar.
  • Using the dashboard — every page in the app, and the Upskill plan.
  • Reference — every CLI command, API endpoint and configuration key.
  • Operations — the runbook, the known failure modes with fixes, and the public deployment.

Press ⌘K (or /) anywhere in the docs to search every page by its title, summary and topics.