Paste a job description. The parser returns the fields you would otherwise type by hand. Company, title, experience level, required skills, salary range, and the keywords an ATS will match against. Each one lands in the correct field in the tracker, so you are not re-reading the same posting three times to remember what it asked for.
3 steps. Each one runs in seconds. Scroll in order.
Copy the full job description out of LinkedIn, Indeed, Glassdoor, or a company careers page and drop it into the box. No character limit on free or paid plans.
The posting goes to a fast model tuned with a tight extraction prompt. Output is a strict JSON shape: company, title, experience level, required skills, nice-to-have, salary range if listed, ATS keywords.
The parsed record populates the job form in one pass. You can edit anything before saving. The full raw JD stays attached to the record so later steps (tailoring, keyword match, rejection analysis) can read it.
One real example, same input reshaped two ways.
Stripe is hiring a Senior Frontend Engineer on the Payments team. You will partner with product and design to turn Figma prototypes into shippable UI using React, Next.js, TypeScript, and our internal design system. You will care about Core Web Vitals, WCAG 2.1 AA, and writing Jest + Playwright tests for everything you ship. Experience in fintech or a regulated environment is a plus. Salary range: 190k to 240k plus equity.
{
"company_name": "Stripe",
"job_title": "Senior Frontend Engineer",
"experience_level": "senior",
"key_skills": ["React", "Next.js", "TypeScript"],
"nice_to_have": ["fintech", "regulated environment"],
"salary_range": "$190K to $240K + equity",
"ats_keywords": ["Core Web Vitals", "WCAG 2.1 AA", "Jest", "Playwright", "design system"]
}The mechanics nobody should have to guess at.
Extraction is a classification task, not a creative one. A small, fast model handles it in roughly a second and costs pennies per parse. A larger model would be overkill and slow the paste-to-saved flow.
The prompt enforces a JSON schema so we never have to parse free-form paragraphs out of the model. If a field is not in the posting (say, salary is absent) we return null for that field instead of guessing.
The tailor step does not need the full 2,000-word posting; it needs the 800 tokens that carry the skills and phrasing. The parser produces a summary alongside the structured fields so later tailor calls stay cheap.
The parser is format-agnostic. LinkedIn reels, Indeed bullet lists, company careers pages with three paragraphs of "about us" before the role. All supported. The model is trained to ignore company boilerplate.
Honest limits read as trust signals. Hiding them does the opposite.
Roughly one to two seconds on a typical JD. Very long postings (3,000+ words) can push toward three seconds. Still under the threshold where you feel the wait.
A small, fast model tuned for extraction. A larger general-purpose model would buy us nothing here and cost more. Free and Pro tiers use the same parser.
No soft limit. Very large pastes (over 20,000 characters) get truncated before parsing, but we keep the full text on the record so tailoring still sees everything.
Yes. Everything is a regular form field after parsing. If the model missed the experience level or misread the salary range, type over it before you save the job.
Yes. When you click the extension icon on a LinkedIn, Indeed, or Glassdoor posting, it reads the visible JD, sends it to the same parse endpoint, and saves a pre-filled job in your tracker.
Every saved job gets its own resume, rewritten to mirror the posting. ATS-safe, zero fabrication.
Matched and missing keywords named exactly. No opaque score, no black-box percent.
Anonymized outcome patterns from the whole opt-in network, not just your own history.
Create a free account in under a minute. First job tracked, first tailored resume, and first keyword breakdown all happen inside the onboarding flow.