Looking for a job is a numbers game and a quality game at once. You need volume — applications out the door, follow-ups sent, interviews booked — and you need each one to be genuinely good. Claude Cowork is built for exactly this kind of workload: it's Anthropic's desktop AI agent that runs tasks in an isolated sandbox VM, with access to your files, coordination between sub-agents, plugins, MCP connectors, and recurring Scheduled Tasks.
This guide walks through five practical workflows: tailoring resumes, writing cover letters, optimising your LinkedIn profile, practising interviews, and tracking applications. Every section includes a prompt you can copy and adapt.
Why Claude Cowork suits job hunting
Claude Cowork isn't a chatbot you type at — it's an agent that works alongside you. It runs on macOS (since January 2026) and Windows (since February 10, 2026), and executes tasks in an isolated sandbox VM rather than directly on your machine. That means it can read and write the files you grant it access to, spin up sub-agents to handle different parts of a task, and use plugins and MCP connectors like Google Workspace or Microsoft 365 to reach the tools you already use.
The practical upshot for your job search: it can hold your whole application pipeline in context — your master resume, the job descriptions you're targeting, your notes — and produce tailored, ready-to-review output without you copy-pasting between tabs.
Before you start: create one folder for your job search (for example /JobSearch/ with subfolders for Resumes, Cover-Letters, Applications, and Briefings), and grant Claude Cowork access to it the first time it asks. Everything below assumes that's in place.
1. Tailoring your resume to a job description
The fastest way to get rejected is sending the same generic resume to every role. The fix is simple in theory and tedious in practice: rewrite your resume so the most relevant experience leads, mirror the language of the role, and cut what dilutes the fit. That's precisely the kind of careful, repetitive editing an agent does well.
Save the job description as a file, keep your master resume alongside it, and use a prompt like this:
Read the job description at /JobSearch/Applications/ACME-senior-analyst.md
and my master resume at /JobSearch/Master-Resume.md.
Rewrite my resume for this specific role:
- Keep every fact true — reorder and rephrase, never invent.
- Lead with the experience most relevant to the role's requirements.
- Mirror the job description's language where it honestly fits.
- Cut anything that dilutes the fit, and tighten the wording.
- Save the result to /JobSearch/Applications/ACME-senior-analyst-resume.md.
Then list every change you made and the reason for each.
Two habits make this workflow sing:
- Keep a "brag bank". A running file of accomplishments with numbers ("cut onboarding time 30%", "led a 6-person migration") gives Cowork strong material to draw from for every tailoring pass.
- Ask for the change log. The "list every change" instruction keeps you in control, so nothing slips through that you wouldn't stand behind. Review it like you would any draft from a colleague.
If you work with a particular workflow often, the career-ops skill in our resources directory packages similar resume and application workflows as reusable instructions you can point Cowork at.
2. Cover letters that don't sound like an AI wrote them
Most AI-generated cover letters share a tell: they're fluent, generic, and about nothing in particular. The trick is to make Cowork draft from your specifics, then push it through a humanising pass.
Using my brag bank at /JobSearch/Brag-Bank.md and the job description at
/JobSearch/Applications/ACME-senior-analyst.md, write three different
cover letter drafts for this role. Each should:
- Open with a concrete reason I'm a strong fit, tied to the role.
- Reference one specific achievement from my brag bank that maps to a
requirement in the job description.
- Stay under 250 words and sound like a person wrote it — no corporate
filler, no clichés, no "I am writing to express".
- Close with one line suggesting a next step.
Save all three to /JobSearch/Cover-Letters/ and tell me which one you
think is strongest and why.
Then run the humanising pass:
Take the strongest draft and rewrite it so it reads like it was written
by me, not an AI: vary sentence length, use my natural vocabulary from
my brag bank, remove any phrase that sounds templated, and keep it
conversational but professional.
Three drafts, one pick, one polish — you end up with a letter built from real facts about you, which is more than most candidates send.
3. LinkedIn profile optimisation
Recruiters read your profile before they read your resume. A well-optimised profile is a passive job-search channel that works while you sleep. Cowork can't improve your profile for you — but it can do the heavy drafting and keyword work.
Here is my current LinkedIn headline, about section, and experience
bullets: [paste them here, or point Cowork at a file].
I'm targeting roles like [senior product manager at SaaS companies].
Rewrite:
1. Three headline options that include my target role keywords.
2. An about section (max 3 short paragraphs) that leads with my
strongest outcome, states what I do, and names the problems I solve.
3. My experience bullets tightened so each starts with an action verb
and includes a number where possible.
Flag any phrasing that sounds like marketing fluff and suggest
plain-English alternatives.
A couple of notes:
- You publish, it drafts. Cowork prepares the text; pasting it into LinkedIn is yours to do, so nothing goes live unread.
- Pull in your notes. If you keep career notes in Google Docs or Microsoft 365, the MCP connectors let Cowork read them directly instead of you copying text around. Developers can also hook up the official GitHub connector to turn public repositories into a tidy portfolio summary for their profile.
4. Interview practice with sub-agents
This is where Claude Cowork earns its keep. Its sub-agent coordination lets one agent play the interviewer, another play a tough technical panellist, and a third score your answers — all in one session.
Run a 30-minute mock interview for the Senior Analyst role at ACME
(see /JobSearch/Applications/ACME-senior-analyst.md).
Use sub-agents:
- One acts as the hiring manager asking behavioural questions.
- One acts as a sharp, sceptical technical interviewer.
- One scores my answers against the job description's requirements.
Ask me one question at a time and wait for my answer. After each answer,
give brief feedback: what landed, what to tighten, and a stronger way
to structure it. At the end, give me an overall score and the three
weakest areas to work on.
When you actually land the interview, have it research the company too:
I have an interview with ACME on Friday. Build a one-page briefing at
/JobSearch/ACME-briefing.md: recent product launches and news, the
role's likely priorities, the interviewer's background if public, and
three sharp questions I could ask them.
Answer honestly in the mock — the scoring is only as useful as the answers you give it. Practise the weak areas it flags, then run a second round focused on those.
5. Tracking applications with a spreadsheet + Scheduled Task
Applications have a way of going quiet. You send five, hear back from two, and lose track of the rest. Fix that with a tracker spreadsheet and a recurring Scheduled Task — Claude Cowork's scheduled automation, available since February 2026 — to keep it honest.
Create a job application tracker at /JobSearch/Tracker.csv with columns:
Company, Role, Source, Date Applied, Status, Next Step, Next Step Date,
Notes.
Then import the five applications currently in /JobSearch/Applications/.
Finally, set up a Scheduled Task that runs every Monday at 9am to:
- Review the tracker and flag applications with no update in 7 days.
- Propose the next action for each flagged application (follow-up email,
new contact, or close it out).
- Summarise the week's pipeline in a short report saved to
/JobSearch/Weekly-Update.md.
If you connect Google Workspace or Microsoft 365 via MCP, the tracker can live in Google Sheets or Excel instead of a local CSV, so it syncs everywhere. For the details of scheduling, check our Scheduled Tasks guide.
6. Keeping your search organised (and honest)
A few principles to keep the whole pipeline healthy:
- One folder, one source of truth. Everything lives under
/JobSearch/, so Cowork can cross-reference your brag bank, resumes, and briefings in one pass. - You send, it prepares. Drafting, tailoring, and tracking are all fair game; actually submitting applications and sending messages should stay yours, so nothing goes out unread.
- Review everything. The sandbox keeps Cowork's work contained and safe, but you're the one accountable for what leaves your machine.
- Know your plan. Cowork is included with Claude Pro ($20/mo) and Max ($100/mo) — see Pricing for details. Heavier interview-prep sessions and large research briefings use more capacity, so plan accordingly.
Job-Hunt Power Pack
If you'd rather not build these workflows from scratch, we're putting together the Job-Hunt Power Pack: a downloadable pack of 20 copy-paste prompts and ready-made workflows for Claude Cowork, covering resume tailoring, cover letters, LinkedIn optimisation, mock interviews, application tracking, and follow-up sequences.
$19 — launching soon. Join the waitlist →
This guide is community-maintained and based on publicly available information. Always refer to Anthropic's official documentation for the latest on Claude Cowork's features, availability, and limits.