01 / 0401
Install the Chrome extension
Sixty-second install from the Chrome Web Store. Trackr asks for read-only Gmail scope — no send, no delete — and authenticates against the dashboard via a single OAuth round-trip.
Index
03Live· 2025· Product Design + Engineering
Exploring AI-augmented job search
Trackr
Parses job application emails with 95%+ accuracy using Claude, automatically organizing your entire job search into a kanban workflow — no manual entry required, even though every company formats their confirmations differently.
0%+
Parse accuracy
0
Pipeline stages
0s
Per-application capture
Live
Production status
01Key Insight
WHY STRUCTURED PROMPTING BEATS RULES
Parsing job emails is an adversarial problem — every company formats confirmations differently. The naive approach is per-company regex rules that break the moment a recruiter rewords their template. Trackr uses Claude with a strict output schema instead: the parser describes what fields it needs, lets the model figure out where they live, and validates the result. 95%+ accuracy without a single company-specific rule.
02The Problem
Why this exists.
Most candidates running a real job search end up in one of two failure modes. Either they're typing application status into a spreadsheet that rots within two weeks, or they've adopted a CRM that was designed for sales pipelines and now spend more time grooming records than interviewing. Neither tool tells you the only things that matter at 9pm on a Sunday: what's overdue, who hasn't responded, and what your actual response rate looks like.
The tools also miss the structure of the data. A job posting URL contains a company, a role, a stage signal, a contact, and a deadline — all of it strict, all of it parseable. Asking a candidate to type those fields by hand is exactly the kind of clerical work software should have eliminated a decade ago.
03The Approach
How it works.
Trackr is opinionated. Every interaction is shaped around three questions: what's my next action, what's overdue, and what's my response rate trending toward. There's no Kanban-archaeology view, no retroactive analytics dashboard, no notes module that nobody uses. The pipeline is five stages — Applied, Screen, Onsite, Offer, Closed — and the home view is a sorted list of next actions with due dates.
The data layer is built around an AI parser. Paste a job-posting URL, forward a recruiter email, or drop in a manual note, and a Claude-powered parser extracts company, role, stage signal, next action, and due date in roughly 12 seconds. The candidate spends time interviewing instead of typing — which is, after all, the whole point.
Response-rate analytics are the second pillar. Every candidate's response rate is benchmarked against their own historical baseline, not a global average — so the metric stays honest as the search shifts across roles, geographies, and seniority levels.
04USER JOURNEY
From inbox to dashboard in four steps.
02 / 0402
Gmail scans for application signals
The extension watches for inbound mail that looks like a job confirmation, recruiter outreach, or interview scheduling — pattern-matched, then handed to the parser. No active forwarding required.
03 / 0403
Claude parses into structured fields
Each candidate email becomes a typed application record: company, role, stage, next action, due date. The schema is enforced at parse time so the dashboard never sees a malformed row.
04 / 0404
Dashboard shows next actions
Everything sorts into a queue keyed on due date. The candidate opens trackrjobs.com once a day and sees a list of decisions — not a spreadsheet they have to grep.
05Features
What it does.
01 / 04Capture
AI-parsed inputs from anywhere
Paste a job-posting URL, forward a recruiter email, or drop in a one-line note — Trackr parses it into a structured application record in about 12 seconds, including company, role, stage signal, next action, and due date. The friction of logging an application drops to roughly zero, which is the only setting at which candidates actually keep tracking.
02 / 04Pipeline
Five stages, ruthlessly enforced
Applied → Screen → Onsite → Offer → Closed. No custom stages, no sub-stages, no opinions about how to model your particular search. The constraint is the feature: every candidate sees the same pipeline shape, which means analytics work, advice transfers, and the home view stays scannable in one glance.
03 / 04Signals
Honest response-rate benchmarking
Trackr tracks response rate against the candidate's own rolling baseline, not a global one. A senior staff search and a new-grad search are different problems; comparing them to each other or to an industry average produces noise. Per-candidate baselines surface real signal — when a strategy is working, you see it within a week.
04 / 04Cadence
Next-action queue, not a Kanban board
The home view is a sorted list of due-and-overdue next actions, not a Kanban. Kanban boards optimize for visualizing flow; candidates need a tool that optimizes for the next thing they have to do. Everything else — pipeline view, history, search — is reachable in one click, but the default is a queue.
06Impact
What it produced.
PARSE ACCURACY
0%+Across LinkedIn, Greenhouse, Lever, and recruiter cold emails — measured on a corpus of real candidate inboxes. No company-specific rules required.
PRODUCTION
LiveShipped to trackrjobs.com with active candidates running real searches through the pipeline. Available in the Chrome Web Store as a verified extension.
EMAIL → RECORD
~12sEnd-to-end latency from forwarded email to structured application card on the dashboard, including Claude inference and database write.
← Back to Index
© 2026