Hire with the
candidate's consent.
A modern ATS that respects the people on the other side. Post a role, receive structured applications parsed by Claude, and review with a clean three-panel diff: raw resume → AI output → validated profile. Every candidate opts in granularly and can withdraw at any time.
For employers
Post a job, get structured applications. AI parses resumes into hiring-relevant fields; your team reviews profiles with a full audit of what was extracted and why.
For candidates
Apply to a specific role, or opt into the talent network to be discoverable across openings. You grant each consent individually and can withdraw at any time.
A fairer hiring process — by design.
Most ATSs treat candidates as inputs to be optimized. We start with a different assumption: you're a person, your data belongs to you, and the platform should prove every claim it makes about how that data is used.
You decide what we process.
Six independent consents, all opt-in. Optional ones are genuinely optional and individually revocable from /apply/me. Decline AI extraction and the platform falls back to declared fields. Decline marketing and we won't email you.
Every action is on the record.
Each consent grant or revoke writes a hash-chained ledger row with timestamp, IP, user agent, and the verbatim text shown to you. You can download the entire chain whenever you want — proof, not promises.
No protected-class signals — anywhere.
Race, gender identity, age, disability, veteran status — none of these are fields in our hiring schema. They literally cannot appear in a recruiter export, regardless of any consent flag. The architecture enforces it.
An ATS that's nice to candidates and useful to you.
One link, structured applications
Each job gets a public URL and its own consent-first apply flow. Candidates upload resumes; Claude extracts hiring-relevant fields with per-field confidence and source tags.
Review with receipts
Deep-inspection view shows raw resume, Claude output, and validated schema side by side. Nothing is inferred in a black box — you see exactly what the model said and why.
PII masked by default
Reviewers see redacted name, email, and phone until they click unmask — which writes an audit event with the reason. Protects both candidate privacy and your team from bias risk.
Three ways to find your next role.
Apply to a specific role
Browse open jobs posted directly to SourceHire and apply to one. Your submission is tied to that posting.
Discover roles
Open roles aggregated from licensed partners. Click through to apply on the original poster — we never collect your data on these.
Join the talent network
Introducing
College Plus.
A free career-prep program for students and graduates within the last 3 years. Six lessons on landing a job in 2026, an interactive survey to personalize your dashboard, and optional recruiter intros for entry-level roles. Everything governed by the same opt-in consent model as the rest of the platform.
- Six-lesson course covering the 2026 job market end-to-end
- 2-minute interactive survey to set your goals
- Optional recruiter introductions for entry-level roles
- Free, forever. Withdraw any time at /apply/me
Wire SourceHire into the stack you already have.
Direct connectors for Workday, SAP, Oracle, and BambooHR. Aggregator coverage (Merge + Finch) for 30+ more. Calendar / video / messaging built in. CSV always works.
- Workday
- SAP SuccessFactors
- Oracle HCM
- BambooHR
- Greenhouse
- ADP
- Gusto
- Merge.dev
- Finch
Real applications only.
Every application is run through a heuristic engine the moment it arrives — duplicate emails, throwaway domains, keyboard-mashed local parts, headless browser markers, cross-job resume reuse, and IP/email velocity. Soft signals stack; we don't over-block.
- No raw IPs or user-agents stored — daily-rotating hashes only.
- Score posts to the recruiter dashboard with the firing-rule keys visible.
- Below 50: clean. 50–80: human review queue. 80+: auto-rejected.
- Source+ Earnings only pays for fraud-clear applications.
Throwaway email domains
15+ disposable-mail domains hard-blocked. Plus-aliased + numeric-only local-parts flagged. Keyboard-mashed local parts (low vowel ratio) get an extra strike.
Velocity caps
Free for the people who need it. Honest math for everyone else.
Hiring early-career grads through College+? Posting via Source+? You pay nothing. Need AI features? Pricing is set at a healthy margin over the actual Claude API cost we incur — no enterprise sales theatre.
Post jobs, review applications, and use the manual workflow. AI features off.
- Unlimited job postings
- Up to 3 team seats
- Manual cohort filters
- Audit log + consent ledger
- Standard email support
Privacy rights you can actually exercise.
Two regulators on two continents wrote the law: candidates own their data, and the platforms that handle it owe answers when asked. We built the answers into the product. No lawyer required, no support ticket queue, no “please allow 90 days for response.”
GDPR
EU / UK / EEA residents- Article 15 — Right of access. Get a structured copy of every piece of data we hold about you.
- Article 17 — Right to erasure. Delete your account and every linked artifact, including resume blobs.
- Article 21 — Right to object. Stop us from processing your data for any non-essential purpose.
- Article 22 — Protection from solely-automated decisions. Every adverse hiring action has a human in the loop.
Consent isn't a checkbox. It's the architecture.
Boundaries enforced at the schema, prompt, and database level — not just promised in copy.
Opt-in only
Every purpose has its own consent toggle. Optional consents are genuinely optional and independently revocable.
Tamper-evident ledger
Every grant and revocation lands in an append-only, hash-chained ledger. You can request your full record any time.
AI extracts, humans decide
Claude normalizes declared facts into structured fields. It does not infer protected attributes — those fields don’t exist in our schema.
Frequently asked
- How is SourceHire different from a standard ATS?
- Most ATSs are built for the recruiter first and layer privacy on as a compliance check. We built the candidate-facing consent flow first — granular toggles, hash-chained ledger, first-class erasure — and the recruiter UI on top. Claude adds structured extraction, not inferred attributes.
- How does Claude process applications?
- Resumes and cover letters are sent to Claude with a versioned prompt that extracts hiring-relevant fields only — experience, education, skills, certifications, preferences. The prompt explicitly forbids inferring race, religion, gender, disability, age, or similar attributes. Our schema has no slots for those fields, and a post-extraction scanner rejects any that slip through.
- Can I use this for just one role, or do I have to run a talent network?
- Either. Post a single job, get applications for that one role. Or run both — candidates who opt in to the network are discoverable across all current and future postings.
- What happens to candidate data if we stop using SourceHire?
- You can export structured profiles in JSON or CSV via the Migration Hub. On account deletion, all candidate data is erased on the platform; the consent ledger retains tombstone entries that prove the erasure happened.
Hire with the receipts.
Post your first role today. Applications flow through a consent-first pipeline by default — audit log on, PII masked, ledger chained.
