Job Search Claude Prompts: 15 Copy-Paste Templates (2026)

Every job search Claude prompt you find online has the same weak spot: it asks Claude to write, but never gives Claude real data about the roles, the company or the person who hires. This guide gives you 15 copy-paste templates for the whole hunt, and marks which ones need live data and which work in a plain chat.

These are templates, not benchmarked results. They follow the documented SyncGTM MCP parameters and standard prompt-writing practice, so run each on one role first and adjust the wording. This post goes deep on prompts. For the end-to-end workflow, read the pillar guide to job search with Claude.

Which Prompts Need Live MCP Data

The split matters. A writing prompt works the same in any chat. A prompt about current jobs, company size or a person's email only works if Claude can call a tool. Here is the full map.

Set Up Before You Paste

For the ten plain prompts you need nothing but a Claude chat. For the five data prompts, connect SyncGTM's MCP once.

• Claude app: Settings, Connectors, Add custom connector, enter https://api.syncgtm.com/mcp, then authorize in the browser. Custom connectors need a paid Claude plan.
• Claude Code: run the command below, then /mcp to authenticate.

claude mcp add --transport http syncgtm https://api.syncgtm.com/mcp

Test it with a free call: "Check my SyncGTM credit balance." That runs check_credits. The LinkedIn scraper MCP guide covers the connection in more detail.

How to Write a Strict Job Search Claude Prompt

Anthropic's own prompt engineering guide recommends being clear and direct and giving the model the context it needs. For a job hunt, that becomes five habits you will see in every strict prompt below:

1. Name the tool or the source."Call search_linkedin_job_openings" or "using only the text I pasted".
2. Paste the real inputs. Resume, posting, notes. Remove your phone number and home address first.
3. Cap the output. Word counts, table columns, a limit on results so credits stay predictable.
4. Ban invention."Do not add skills, numbers or employers that are not in my resume."
5. Add a stop."Show me the table before calling another tool" or "Do not send".

Stage 1: Job Search Claude Prompt Templates for Finding and Scoring Roles

Claude cannot browse LinkedIn on its own, so this is where live data matters most. These three prompts turn a vague wish into a ranked list.

1. Pull live roles

Needs live MCP data search_linkedin_job_openings · 0.3 credits per job returned

Paste in: Target titles, locations, remote preference, seniority, how recent.

Find me some marketing jobs.

Call search_linkedin_job_openings with:
- title: ["Product Marketing Manager", "PMM"]
- locations: ["Canada"]
- work_type: ["remote", "hybrid"]
- experience_level: ["mid-senior"]
- published_at: "r604800" (past week)
- exclude_recruiting_agencies: true
- save_only_unique_items: true
- max_results: 25

Return a table: title, company, location, posted date, job URL.
Stop and show me the table before calling any other tool.

Why it works: The vague version has no tool call, so Claude answers from memory with listings that may not exist. The strict version names the tool, every filter and a stopping point, which also caps spend at roughly 7.5 credits.

2. Score fit against your resume

MCP data + Claude judgement search_linkedin_job_openings (resume_keywords) · 0.3 credits per job if you re-run the search

Paste in: Resume text (no phone or home address) and the keyword list you want matched.

Which of these jobs fit me?

My resume: [paste resume text].

Using only the 25 roles already in this chat, rank them best to worst fit.
For each give: score out of 10, two reasons it fits (quote my resume), one
real gap, and whether a salary is stated. Do not credit me with any skill
that is not written in my resume. If the posting is thin, say "not enough
detail" instead of guessing.

Why it works: Ranking from the results already in the chat costs no extra credits. If you want the tool to score too, re-run prompt 1 with resume_keywords as an array of keyword and aliases pairs, such as Salesforce with alias SFDC.

3. Decode a job description

Plain Claude

Paste in: The full job description text for one role.

Summarize this job description.

Job description: [paste].

Return four lists:
1. Must-haves (stated as required, with the exact phrase)
2. Nice-to-haves
3. Signals about the real problem this hire solves (read between the lines)
4. Red flags (vague scope, 5+ jobs in one, unrealistic years of experience)

Then give three questions I should ask the hiring manager. Quote the
posting for every claim. Do not infer anything the text does not support.

Why it works: A summary repeats the posting. The four lists tell you what to emphasise in your resume and whether the role is worth chasing.

Stage 2: Research the Company

Before you spend credits on contacts or hours on an application, check that the company is real, funded and still hiring.

4. Company snapshot

Needs live MCP data enrich_organization · 2 credits per company

Paste in: The company website domain (enrich_organization takes a domain, not a LinkedIn URL).

Tell me about Northwind Analytics.

For my top 5 roles, call enrich_organization using each company's
website domain. Return a table: company, industry, employee count, funding
stage and any technologies listed. Mark any field the tool did not return as
"not returned". Do not fill gaps from memory. Tell me the total credits used.

Why it works: Claude's memory of a mid-size company is often out of date. The strict version forces fields from the tool and labels the gaps. At 2 credits each, run it only on your shortlist.

5. Hiring momentum check

Needs live MCP data job_openings_growth_rate, head_count_growth_rate · 3 credits and 2 credits per company

Paste in: The company LinkedIn page URL (the identifier).

Is this company growing?

For my top 3 companies only, call head_count_growth_rate and
job_openings_growth_rate using each company's LinkedIn page URL as the
identifier. First tell me the credit total (5 per company) and wait for my
OK. Then report the trend for each and say whether it suggests a funded,
growing team, a steady team or a shrinking one. Flag any company where
openings are rising while headcount is falling.

Why it works: Both calls are among the pricier company tools, so the strict prompt gates the spend. Rising openings with flat headcount can mean churn rather than growth.

6. Culture and risk read from material you paste

Plain Claude

Paste in: The careers page, the About page, one recent press item and any Glassdoor-style reviews you copied yourself.

Is this a good company to work for?

Material I pasted: [careers page, about page, press, reviews].

Using only this material, give me: what the company says about itself,
what employees repeatedly complain about or praise (count the mentions),
and three claims I should verify in an interview. Label each point
"stated by company" or "stated by reviewers". If the material is too thin
to judge, say so.

Why it works: Claude cannot read private review sites and will invent a culture if you let it. Restricting it to pasted text keeps every point traceable.

Stage 3: Find the Hiring Manager and a Verified Email

This is the stage most prompt lists skip, because it needs data Claude does not have. Treat the output as a shortlist you confirm yourself.

7. Hiring manager shortlist

Needs live MCP data find_people · 0.2 credits per person returned

Paste in: Company name or LinkedIn company URL, the role title, and the function it sits in.

Who is the hiring manager for this job?

For the Product Marketing Manager role at Northwind Analytics, call
find_people with company_linkedin_urls: [paste URL], job_titles:
["Director of Product Marketing", "VP Marketing", "Head of Product Marketing"],
job_levels: ["Director", "VP"], limit: 3.

Return name, title and LinkedIn URL. Say which one most likely owns the
role and why. Label this a guess. Do not state that anyone "is the hiring
manager".

Why it works: No tool can read a company's org chart as a fact. The strict prompt asks for a ranked guess with reasoning, which is what you can honestly use.

8. Verified work email

Needs live MCP data find_work_email, verify_email · 1 credit per find (misses not billed), 0.1 per verify

Paste in: The person's LinkedIn profile URL and the company domain.

Get me their email.

Call find_work_email for [LinkedIn URL] with domain [company domain].
If an address comes back, call verify_email on it. Show: name, email,
verification result. Discard any address that is not confirmed deliverable.
If nothing is found, say so and do not guess a pattern like
first.last@domain. Do not send or draft anything yet.

Why it works: A guessed address that bounces looks careless. Verification costs a tenth of a credit, and the no-guessing rule saves you from the most common mistake.

Stage 4: Tailor Your Resume and Cover Letter

These two are plain Claude work. The only inputs that matter are your real resume and the posting, and the main risk is invention.

9. Resume tailoring without invention

Plain Claude

Paste in: Your master resume and the must-haves list from prompt 3.

Rewrite my resume for this job.

Master resume: [paste]. Must-haves from the posting: [paste].

Suggest edits only. For each change show: the original bullet, the
rewritten bullet, and which must-have it supports. Rules:
- Reorder and reword; never add a skill, tool, employer, date or number
  that is not in the master resume.
- If a must-have has no support, list it under "gaps" instead of stretching.
- Keep every bullet under 25 words, starting with a verb.
- Keep the one-page limit.

Why it works: Asking for suggestions with a source for each one makes the changes auditable. You accept or reject each line instead of trusting a rewritten document.

10. Cover letter from two facts

Plain Claude

Paste in: The posting, one specific thing you noticed about the company, and two achievements with real numbers.

Write a cover letter.

Posting: [paste]. Something specific I noticed: [one sentence].
My two proof points: [achievement with number] and [achievement with number].

Write a cover letter under 180 words, three short paragraphs: why this
company, what I did that maps to the role, a plain closing line. No
"passionate", "synergy" or "leverage". Use only the facts above. Output
the text only.

Why it works: Two real proof points beat a long tour of your career. The word ban and length cap remove the generic phrasing that readers skim past.

Stage 5: Cold Note and Follow-Ups

One person per role, a few notes a day, every draft read and sent by you. Guardrails come first because this is the stage where tools get misused.

11. The first note

Plain Claude

Paste in: The verified contact, the role, one line from the posting, one proof point from your resume.

Write an email to the hiring manager.

Contact: [name, title]. Role: [title]. From the posting: [one line].
My proof point: [one fact from my resume].

Draft one email, max 90 words, plain text. Line 1 names the role and the
posting detail. Line 2 gives the proof point. Line 3 is one low-pressure
ask. Subject under 7 words. Add one line saying they can tell me to stop.
Do not send. Output the draft for my review.

Why it works: The structure forces specificity. The opt-out line and the no-send rule keep you on the right side of both etiquette and email law.

12. Follow-up after a week

Plain Claude

Paste in: The original note and the date you sent it.

Write a follow up.

My first note, sent 7 days ago: [paste].

Write a follow-up under 50 words that adds one new piece of value (a
relevant observation about their product, or a short result I can share),
not "just checking in". Keep the same subject thread. One soft ask.

Why it works: Follow-ups fail when they only repeat the ask. A new, specific detail gives the reader a reason to reply.

13. The graceful close

Plain Claude

Paste in: The thread so far.

Write another follow up.

Thread: [paste both earlier notes].

Write a final note under 40 words that thanks them, says I will stop
reaching out here, and leaves the door open (one line on what I would
still welcome, such as a pointer to the right person). No guilt, no
pressure. This is my last message to this person about this role.

Why it works: Two notes and a close is the ceiling. Stopping on your own terms keeps the contact warm and protects your reputation.

Stage 6: Interview Prep

Once a conversation is booked, the best prompts reuse everything you gathered in the earlier stages.

14. Questions and a story bank

Plain Claude

Paste in: The posting, your resume, and your company notes from prompts 4 to 6.

Give me interview questions.

Posting: [paste]. Resume: [paste]. Company notes: [paste].

1. List the 10 questions most likely for this role, split into role skills,
   behavioural and company-specific.
2. For each, name which resume experience I should use as the answer.
3. For the 5 behavioural ones, outline a STAR answer using only facts in my
   resume, and mark any place where I need to supply a real number.
Do not write answers that claim results I have not stated.

Why it works: Linking each question to a real experience builds a story bank you can actually deliver. The marker for missing numbers stops Claude from filling them in.

15. Mock round and your questions back

Plain Claude

Paste in: The job description, your company notes and the interviewer's title.

Run a mock interview.

Act as the interviewer: a [title] at [company]. Ask one question at a
time and wait for my answer. After each answer give: a score out of 5, what
was strong, what was vague, and a tighter version of my answer using only
facts I gave. After 6 questions, suggest 5 sharp questions I can ask them
and draft a three-line thank-you note I will edit before sending.

Why it works: One question at a time creates a real rehearsal. Ending with questions to ask and a thank-you draft covers the parts candidates usually forget.

What the MCP Prompts Cost

Plain prompts cost nothing beyond your Claude plan. The data prompts spend SyncGTM credits per call. One pass over a three-company shortlist looks like this:

Prompt 5 is the priciest line, which is why its strict version asks for confirmation first. Skip it for companies you already trust. Rates can change, so confirm in the credits estimator and see plans on the pricing page. For writing the notes themselves, our Claude cold email guide goes further, and the email finder MCP post explains the lookup behind prompt 8.

Guardrails for Every Prompt

• Review before sending. Every draft passes through you. Never let Claude send unattended.
• No auto-apply.Bulk-submitted applications carry wrong details and can break a job board's rules.
• No mass emailing. These prompts target one person per role. Blasting a template is spam and can breach laws such as CAN-SPAM (see the FTC compliance guide) and GDPR.
• Check every claim. Compare each rewritten line with your original resume. If you cannot defend it in an interview, remove it.
• Hiring manager lookups are guesses.Confirm on the person's profile before you address them as the hiring manager.

All example outputs on this site use placeholder companies, people and .example addresses. Packaging these prompts into a reusable command is the next step: see the pillar guide for the full loop, and the GTM engineer jobs map if you are hunting in revenue roles.

Conclusion

A strong job search Claude prompt is specific, grounded in text you pasted and bounded by rules. Ten of these templates need only a chat window. The five that find roles, companies and people need live data, and that is where a connected MCP changes the answer from plausible to real.

Start with prompt 1 for the one role you want most, walk a single company through stages 2 and 3, and only then scale up. You can start with SyncGTM for free.