Job Lobster started as a two-month side build in the spring of 2026, after one too many job-search tools that flattered candidates instead of leveling with them. We wanted a workbench that tells the truth about fit, does the grunt work, and leaves the decisions to you.
Existing tools score you 87/100 on roles you have no business applying to — then upsell you a “rewrite” service to get you over the line. That inflation costs time, not tools. A candidate doing a serious search spends months. Every inflated 9/10 is a week spent on a role that was always a 5.
Claude Opus 4.7 + Sonnet 4.6 do the drafting, the matching, the research, and the agentic follow-ups. We don’t maintain a scraped database of jobs — we pull directly from 11 applicant-tracking systems the moment a role posts. That means we can stay private, small, and obsessed with one user: the person running their own search.
The whole codebase is public at github.com/AdamJurgens/JobLobster. No dark patterns, no growth loops, no “claim your free trial by inviting five friends.” If the product is good, you stay. If it’s not, you leave with every byte of your data (export button, always on).
A capped 5.0 with a missing must-have is more useful than an inflated 8.5 with a hidden catch. We show the cap, the rule that fired, and what would change it.
Everything you put in is exportable as a single JSON blob. Delete-my-data wipes per-user tables, per-user files, and optionally your account record — gated behind a typed-confirmation guard. No training on your data, no reselling, no sharing.
We take a position on resume bullet structure, on gap-ceiling rules, on what “a good cover letter” looks like. We cite the reasoning and let you override.
No enterprise add-ons, no platform pivots, no sudden direction-change every two months. We ship small slices, write the reasoning into the commit log, and keep the changelog public.
Early access is free. Sign in with a magic link, upload your resume, see your honest score in under a minute.
Get started →