n8n + n8n: AI Job Application — Tailored Resume & Cover Letter in 60 Seconds
You want a job change but every application takes 40 minutes: read the posting, rewrite your resume to mirror its language, then write a cover letter that doesn't sound like the last 12. So you apply
You want a job change but every application takes 40 minutes: read the posting, rewrite your resume to mirror its language, then write a cover letter that doesn't sound like the last 12. So you apply to three roles instead of thirty, and the funnel dies from low volume. The bottleneck isn't your qualifications — it's the manual rewrite tax on every single application.
The real problem: generic applications lose to tailored ones
Recruiters and ATS software both reward specificity. An ATS scores your resume against the exact keywords in the job description — "Salesforce administration" matches, "CRM experience" often doesn't. A human recruiter spends six seconds deciding whether you read the posting or blasted a template. Both signals demand tailoring, and tailoring by hand doesn't scale past a handful of applications per week.
The usual workarounds fail. A single "master resume" gets filtered out because it never mirrors the posting's vocabulary. Manually editing each one burns the time that should go into interview prep. Copy-pasting your resume into ChatGPT for every role means re-explaining your background, re-pasting the job text, and babysitting the output — you've moved the work, not removed it. What you need is a pipeline: paste a URL, get back two documents already aligned to that role.
The solution: a URL-in, documents-out n8n workflow
The template does one thing well. You give it the job posting URL. It scrapes the page, extracts the role's requirements and the company's language, then generates a resume rewritten around those exact requirements plus a cover letter that opens with something specific to that company — not "I am writing to express my interest," but a first line that names their product, mission, or a detail only someone who read the page would know.
Under the hood it's a linear n8n flow: an HTTP request pulls the posting, a parsing step isolates the job description from the page's navigation and footer, an AI model rewrites your stored resume against that description, a second AI step drafts the cover letter, and the outputs land wherever you want them — Google Docs, an email to yourself, or a Notion page. No manual step between paste and finished draft.
Step-by-step setup in n8n
Here's how the workflow is wired. If you buy the template you import it and fill in credentials; if you're building it yourself, these are the nodes and the config that matter.
1. Trigger — how the job URL gets in. Use a Form Trigger node with a single required field named job_url (type: URL). This gives you a hosted form you can bookmark — paste, submit, done. If you'd rather fire it from a browser bookmarklet or another tool, swap in a Webhook node set to POST and read {{ $json.body.job_url }} downstream.
2. Fetch the posting — HTTP Request node. Method GET, URL set to {{ $json.job_url }}. Under Options enable "Response → Include Response Headers and Status" so you can catch a 403. Add a realistic User-Agent header (many job boards block default agents). For JavaScript-heavy pages like LinkedIn or Greenhouse that render client-side, route through a rendering service — a HTTP Request to a ScrapingBee/Browserless endpoint, or n8n's community Puppeteer node — so you get the fully painted HTML.
3. Strip the HTML to text. Add a Code node (JavaScript) that takes $input.first().json.data, removes <script> and <style> blocks with a regex, strips remaining tags, and collapses whitespace. Cap the output at roughly 6,000 characters — job pages carry a lot of boilerplate, and you only want the description, responsibilities, and requirements to reach the model. Clean input here directly improves the tailoring quality.
4. Tailor the resume — AI model node. Use the Anthropic Chat Model node (or an OpenAI node — the flow is model-agnostic). This is the core step, so use a capable model: set the model to claude-sonnet-5 for a strong quality-to-cost ratio, or claude-opus-4-8 when you want the sharpest rewrite. Give it a system prompt like: "You are an expert resume writer. Rewrite the candidate's resume below to align with the target job description. Mirror the job's terminology where the candidate genuinely has the experience. Never invent experience. Return clean, ATS-friendly text." Pass two things in the user message: your base resume (store it once in a Set node at the top of the flow, or pull it from Google Docs) and the cleaned job text from step 3. Set maxTokens to around 2,000 and temperature to 0.4 — low enough to stay grounded, high enough to rephrase naturally.
5. Write the cover letter — second AI node. Chain a second Anthropic Chat Model node. Its prompt instructs the model to open with one specific sentence about the company drawn from the posting, then connect the candidate's two most relevant achievements to the role's top requirements, and close with a direct call to action. Feed it both the job text and the freshly tailored resume so the letter and resume tell the same story. Keep it under 300 words — recruiters don't read long ones.
6. Deliver the output. End with a Google Docs node (create document, title {{ company }} — Resume) plus a second doc for the letter, or a Gmail / Send Email node that mails both to you as the body. For a paper trail, append a row to a Google Sheets node logging the URL, company, and timestamp so you never apply to the same role twice.
Why this beats doing it by hand
The obvious win is speed: paste to finished draft in under a minute versus 40 minutes of manual editing. But the compounding win is volume with quality intact. When each tailored application costs you one paste, applying to 25 roles a week becomes realistic, and every one of them still mirrors its posting — so your ATS pass-rate and recruiter response-rate stay high instead of collapsing under batch-and-blast generics.
It's also consistent. Your best framing of a given achievement doesn't depend on whether you're writing the application at 9am or 11pm. And because the base resume lives in one Set node or Google Doc, updating a new accomplishment once propagates to every future application automatically. For an ops team running this for candidates, or a founder hiring while job-hunting, that single source of truth is the difference between a system and a pile of Word files.
Common pitfalls and how to avoid them
Blocked scrapes. LinkedIn, Indeed, and Workday aggressively block bots. If your HTTP Request returns a 403 or an empty body, that's the cause — route through a rendering/proxy service and set a real User-Agent. Test with a Greenhouse or Lever posting first; those are cleaner to scrape while you validate the flow.
Hallucinated experience. The single biggest risk. If your prompt says "make the candidate a great fit," the model will happily invent skills you don't have — and that gets you caught in the interview. Pin every prompt with "Never invent experience; only rephrase what's in the base resume." Treat the output as a first draft you skim, not a document you submit blind.
Dirty input text. If you skip the HTML-cleaning Code node and pass raw markup to the model, half your token budget goes to navigation menus and cookie banners, and the tailoring degrades. The cleaning step isn't optional — it's what makes the AI focus on the actual requirements.
Token and cost creep. Two AI calls per run is cheap, but an uncapped resume plus a bloated job page can balloon input tokens. Cap the scraped text, cap maxTokens, and pick the model deliberately — a mid-tier model handles most rewrites fine; reserve the top model for senior roles where phrasing matters most.
Same-role duplicates. Without the logging step, you'll re-apply to postings you already hit. The Google Sheets append plus a quick lookup at the top of the flow prevents the awkward double-submission.
Ja construimos isso pra voce
Nao comece do zero. O AI Job Application — Tailored Resume & Cover Letter in 60 Seconds e um workflow n8n pronto para instalar — conecta suas ferramentas em minutos, sem codigo.
Instalar por $29.0 →
