About GitFinder

A sourcing tool for technical recruiters and engineering leaders who are hiring. You paste a job description; it derives a GitHub query plan, finds the repositories where that kind of work is happening, mines their contributors, screens those people on evidence rather than self-description, ranks each one against your specific role with a written reason, and resolves a usable email address from public data.

The problem it solves is narrow and real: GitHub holds the best public record of what engineers have actually built, and its search is not designed for finding people. Searching GitHub’s user index only ever tells you what someone typed into their profile. The useful signal is in repositories and commit history, and getting at it means several hundred API calls per role against two separate rate limits. That is a software problem, and this is the software.

It is not an attempt to replace a recruiter’s judgement. The ranking exists so you know whose code to read first, and every score comes with an explanation specifically so you can disagree with it.

GitFinder was created by Rahul Vishwakarma, who designed and wrote it. It is operated commercially by WorkOnward, Inc. — the entity that bills for the subscription, controls the data, and is legally responsible for the service.

That split is worth stating plainly, because it is what the footer, the Terms and the structured data on every page all say: the author is a person, the publisher is a company.

Full operator details, so there is no ambiguity about who you are dealing with:

• WorkOnward, Inc.
• 124 East 14th Street, New York, NY 10003, United States
• info@workonward.com · +1 347-537-9363
• www.workonward.com

GitFinder is one of several products WorkOnward operates in the hiring space. It is a small, self-funded team, which is why the pricing is a single monthly plan with no sales call attached.

Four stages, in this order, because each one narrows the pool cheaply:

1. Read the role. A model turns the job description into a structured query plan — one or two languages, a few ecosystem topics, a deliberately low star floor, an activity window, and a location only if the role genuinely requires one.
2. Discover. Repository search finds the projects matching that plan; their contributor lists become the candidate pool. The source repository is carried with each handle, which is both outreach context and the thing that makes contact resolution cheap.
3. Screen and rank. Bots, organizations, drive-by contributors and dormant accounts are dropped. The survivors are scored individually on language depth, recent activity and project relevance, each with a written reason.
4. Resolve contact details. A four-source cascade over public data — profile email field, commit metadata, own repositories, GPG key user IDs — ordered by yield per API call and exiting at the first usable address.

The full method, including the parts you can do by hand, is written up in how to find and source developers on GitHub and how to find a developer’s email address. We publish it because the method is not the moat; doing it reliably inside two rate limits is.

Every number on this site comes from the production system, and the population it was measured on is always stated. That matters more than the number: a hit rate without a population is marketing.

Two figures we publish that are not flattering, because they set honest expectations. Around 29% of the email addresses in public commit metadata are @users.noreply.github.comprivacy artifacts, which we discard rather than decode — so nobody should believe a tool claiming near-total email coverage. And GitHub’s 5,000-requests-per-hour budget against roughly 500 calls per search means about nine full searches an hour per token, which is the real reason daily limits exist.

We also correct a widely-repeated claim. The technique most guides still recommend for pulling emails from GitHub — reading payload.commits[].author.email from the public events feed — no longer works; GitHub removed the commits array from PushEvent payloads, verified live against our own token. Tools built on it fail silently rather than loudly.

These are design decisions, not missing features, and each one costs us something:

• It never works around a privacy setting. A @users.noreply.github.comaddress is someone saying “keep my email private” in the clearest way the platform allows. We reject it outright and never strip the prefix to guess at the real address — even though doing so would visibly improve our coverage number.
• It scrapes nothing behind a login. Public GitHub data only. No LinkedIn, no purchased contact databases.
• It doesn’t send email. We give you an address and context. Bulk sequences from a sourced list are how sourcing tools get everyone blocked, and the personal first line is the part that actually gets replies.
• It isn’t a screening or background-check product. Sourcing signals are for deciding who to talk to. Using compiled third-party data to assess employment eligibility is regulated territory, and our Terms prohibit it.
• A score is never the decision. Rankings are decision support with a stated rationale. We also say out loud that GitHub is a biased sample — public work correlates with free time, and entire industries work in private — so a GitHub-only pipeline should be used to find people, never to rule them out.
• Developer removal is unconditional. If you are an engineer who would rather not be in GitFinder, email info@workonward.com with your GitHub handle. No justification, no account needed, and it is a permanent suppression rather than a delete that the next search undoes. Details in the Privacy Policy.

Briefly, for the engineers who ask — and most of the people this tool finds are engineers:

• Frontend: Next.js App Router with React, server-rendered, which also owns authentication, billing and entitlement.
• Backend: FastAPI with an async GitHub client, streaming progress to the browser over server-sent events so a 49-second search shows its work instead of spinning.
• Data: self-hosted MongoDB. Your searches and lists are scoped to your account and live on our own infrastructure, not a third-party data platform.
• Models:Anthropic’s Claude — one careful pass to read the role, then per-candidate scoring. Job descriptions and public profile summaries go to the model; email addresses, credentials and payment data never do.
• Payments: Stripe, hosted checkout and billing portal. We never see a card number.

The full list of processors, what each one receives, and how long we keep things is in the Privacy Policy — as a table, not prose.

info@workonward.com reaches a person. Use it for billing problems, a search that behaved oddly, a removal request, a higher-volume conversation, or to tell us the tool is wrong about something — especially the last one.

Post: WorkOnward, Inc., 124 East 14th Street, New York, NY 10003, United States. Phone: +1 347-537-9363.