ATSPrep MCP Server

Use this when the user wants a cover letter for a job. Draft a cover letter from the user's resume and the job posting with fixed rules: name, contact details and up to two results from the resume, the job title and top keywords from the posting. Facts that are missing show as [square-bracket] gaps for the user to fill, never as stand-ins. Returns the letter, the gaps and a checklist.

Use this when the user wants to resign and needs a resignation letter or their last working day. Draft a short, professional resignation letter and work out the last working day from the notice period. Missing facts show as [square-bracket] gaps. Returns the letter, the last day and a checklist (notice stated, no complaints, length).

Use this when the user wants their LinkedIn About or summary section reviewed. Check a LinkedIn About section: what shows before "see more" on a phone and on a computer, whether the opening says who they are, results with numbers, buzzwords, a clear next step, and the 2,600-character limit.

Use this when the user wants their existing LinkedIn headline reviewed. Check a LinkedIn headline: length against the 220-character limit, what shows in search results before the cut, whether it leads with a searchable job title, and buzzwords or vague claims to replace.

Use this when the user wants stronger verbs or synonyms for resume wording. Suggest resume action verbs. Pass a weak opener ("responsible for", "helped") for direct swaps and why they read weak, or the work in plain words ("trained new staff", "cut costs") for the matching verb groups with an example bullet. Leave query empty for every group.

Use this when the user wants a few resume bullet points checked without sending the whole resume. Give each resume bullet a score, a verdict and specific fixes with stable codes: weak openers, missing scale or result, numbers that show neither, hard-to-believe claims, first person, vague claims. The same rules and severities as check_resume. Up to 30 bullets of up to 600 characters each; blank items are skipped and reported by index. Rules version 2026-10-31.1.

Use this when the user wants new LinkedIn headline ideas. Write LinkedIn headline options from the user's answers, each in a proven pattern (title and focus, title and result, title and who you help…), with its length and roughly what shows in search before the cut. Uses only the answers given.

Use this when the user wants one resume bullet rewritten or made stronger. Check one resume bullet with the fixed rules (score, verdict, fixes) and suggest up to three stronger rewrites that keep every fact as given. ask_user names the one fact to ask the user for; never invent it. Part of the answer is drafted by an AI model (Gemma 4, with Llama 3.3 70B as a backup, on Cloudflare Workers AI) behind ATSPrep's guard, which drops any figure, name or tool the user's text doesn't contain. If the AI part is unavailable, the rule-based part still comes back and ai.message says why.

Use this when the user is preparing for an interview and wants likely questions, for a specific job or in general. Interview questions to practise, each with what the interviewer is listening for, plus STAR answer tips. With a job_description, questions come from its skills and duties, along with what to research and questions to ask; with resume_text, from the user's own roles. Rule-based.

Use this when the user gives a job posting link instead of its text; then pass the result to the job tools. Read a public job posting from its link and return the title, company, location and full description as plain text, ready for match_job, tailor_resume, keyword_scan or interview_questions. Reads LinkedIn, Greenhouse, Lever, Ashby, SmartRecruiters, Workable, Workday and most company careers pages. Some sites block reading; then ask the user to paste the text.

Use this when the user has only a job posting (no resume yet) and wants its key skills and keywords. List the skills, tools and requirements a job posting asks for, ranked by how hard it leans on each (repetition, position, required or preferred headings), with the high-priority ones flagged. Use match_job to see which a resume already shows. Posting 120–20,000 characters. Rules version 2026-10-31.1.

Use this when the user has a resume and a job posting and asks how well they match, what keywords are missing, or whether to apply. For rewrites, use tailor_resume instead. Compare a resume with a job posting: the posting's keywords found in the resume (with how they matched: exact, normalized, alias or stem, and the matched span), the ones missing with the posting sentence they come from, each term's weight and whether the posting lists it as required or preferred, and weighted overall, required and preferred coverage. Posting 120–20,000 characters. Rules version 2026-10-31.1.

Use this when the user wants to edit their resume in an editor, apply a template, or download it as a PDF or Word file. Make a link that opens this resume in ATSPrep Studio, the free editor, so the user can fix it, pick a template and download a PDF. The text travels in the link's #fragment, which is never sent to a server.

Use this when the user shares a resume (pasted text or a file's text) and wants it checked, scored, reviewed for ATS problems, or wants to know what to fix. Score a resume from 0 to 100 with ATSPrep's fixed rules and list each issue with a stable code, where it is and how to fix it. Also reports how the text parsed (roles, experience and project bullets, education, skills and skill groups, input and scored word counts, with definitions). Role-specific advice, such as seniority signals, is given only when the resume names a target title or you pass target_role or job_description; with job_description the score becomes the fit for that job (0.5 requirements met, 0.3 evidence, 0.2 readability, capped for missing hard requirements or impossible claims) and it also returns job_fit and a keyword match summary. Pass language when the resume isn't in English. Resume up to 30,000 characters. Rules version 2026-10-31.1.

Use this when the user gives an answer to an interview question and wants feedback on it. Review the user's answer to an interview question: a rating, length and speaking time, STAR coverage for stories, filler words, specific checks, the follow-up an interviewer would likely ask, and a stronger version that keeps to the user's own facts. Part of the answer is drafted by an AI model (Gemma 4, with Llama 3.3 70B as a backup, on Cloudflare Workers AI) behind ATSPrep's guard, which drops any figure, name or tool the user's text doesn't contain. If the AI part is unavailable, the rule-based part still comes back and ai.message says why.

Use this when the user asks what jobs or roles they could apply for, or what their resume fits. Read the resume against what postings for each of 60 roles (office, health, retail, hospitality, trades, tech, sales, design and more) usually ask for, phrased as postings in Kenya, the Gulf, India, the UK and the US put it, and return up to 3 roles: how many of those asks the resume shows (a count, never a percent), its strongest ones, the first gaps, whether it matches their own job title, and search links for openings near them. These are typical asks, not a real posting: then ask the user for a real opening and run match_job or tailor_resume on it. Never tailor a resume to the typical asks.

Use this when the user wants a resume summary or professional profile written from their resume. Draft professional summaries from a resume: rule-based drafts built from its title, skills and the bullets that carry a figure, plus AI drafts that use only facts in the resume. Part of the answer is drafted by an AI model (Gemma 4, with Llama 3.3 70B as a backup, on Cloudflare Workers AI) behind ATSPrep's guard, which drops any figure, name or tool the user's text doesn't contain. If the AI part is unavailable, the rule-based part still comes back and ai.message says why.

Use this when the user wants their resume rewritten or tailored for a specific job posting. Tailor a resume to a job posting: the posting's keywords the resume is missing (from match_job's rules) and suggested rewrites of the user's own summary and bullets that bring in the posting's terms only where the existing line already supports them. Each suggestion lists the posting terms it borrowed and confirm_with_user lists terms to ask about, never to add unasked. End the reply with next_step: the link that opens the resume in ATSPrep to check the edits and download. Part of the answer is drafted by an AI model (Gemma 4, with Llama 3.3 70B as a backup, on Cloudflare Workers AI) behind ATSPrep's guard, which drops any figure, name or tool the user's text doesn't contain. If the AI part is unavailable, the rule-based part still comes back and ai.message says why.

Use this when the user asks for resume templates, designs or example resumes for a role. List ATSPrep's resume designs, each with a link that opens Studio in that design, and the example resumes by role, each with a link that opens it in Studio and a preview image.