com.aicandidatehub/mcp

Analyze how a specific JD (by jd_id from the Jobs Module) or a free-text role+location compares to market compensation. Returns market percentiles, the user's position if comp data is on file, and AI insights.

Analyze how well your CV (also called resume — terms interchangeable) profile matches a specific job description. Mirrors the working browser flow at src/app/jobs/page.tsx:461-494: stage 1 calls the gated job-fit BFF /api/analyze/job-fit (which meters the job_fit_analysis quota, then forwards to the Python analyzer), stage 2 persists to Go /api/v2/job-analyses (best-effort). Returns alignment scores (0-100 scale), strengths, weaknesses, recommendations. Enum hints (documented for agent UX, not enforced here — upstream may relax): • level: entry | mid | senior | lead | principal • recommendation: strong-fit | fit | partial-fit | not-fit Returns envelope.meta.score_scale="0-100" and envelope.meta.persisted=<bool>.

Generate a customized cover letter for a specific job description using AI. Uses your CV (also called resume — terms interchangeable) profile as the candidate context. Optionally persists the result to your library. Focus values: 'balanced' | 'cv-focused' | 'job-focused' (DB-enforced). Synonyms accepted; unknown values fall back to 'balanced'. Response includes the resolved focus on envelope.echo.input_focus and the normalized value on envelope.meta.focus_used.

Fetch compensation benchmarks for a role. Two modes: (1) JD-context — pass jd_id from list_jobs/get_job/list_tracked_jobs.jd_id (RECOMMENDED for SG roles; mirrors browser pattern, closest-match relaxation via Gemini applies automatically). (2) Free-text US benchmark — pass role_title + country='US' for a US benchmark via the compensation analyze endpoint (job_id empty → no persistence). For SG roles without a jd_id, call import_job_from_url first to obtain a jd_id with extracted SSOC codes.

Return a short, copy-paste-friendly download URL for a previously-generated CV (also called resume — terms interchangeable). Use this when the original signed URL from rewrite_cv_for_job has expired (1 hour TTL) or got line-wrapped in the chat client. The returned short_url routes through the app's /d/[code] redirect, which re-signs the stored Supabase object on each follow (no DOCX/PDF re-render). Short URL itself is valid for 7 days. IMPORTANT: format is fixed at the time the CV was generated by rewrite_cv_for_job — this tool does NOT re-render across formats. If you need a different format than was originally generated, call rewrite_cv_for_job again with the desired export_format. PDFs generated via MCP are rendered with contentType='standard' and length='3_page'; this is fixed at generation time.

Fetch a single Job Description by id from the user's Jobs Module library. Returns the full parsed JD content (parsed_content, salary, classification codes). This is the canonical lookup used by analyze_job_fit, generate_cover_letter, and rewrite_cv_for_job. For Job Tracker pipeline entries (status/dates/notes), use get_tracked_job.

Fetch the authenticated user's candidate profile including about, experience, education, and skills. Safe for display to the user.

Returns MCP server version, commit SHA, environment, and deployment timestamp. Useful for bug reports and debugging which version of the MCP is responding. No authentication required.

Fetch a specific job from the Job Tracker pipeline by ID. Returns the full tracker row (status, dates, jd_id). The job_tracker table does not store free-text notes — activity history lives on the activities endpoint.

Import a job listing from a URL. The backend auto-detects LinkedIn vs other sources. Parses the JD, persists it to the Jobs Module library, and returns the new jd_id for use with analyze_compensation_for_job, analyze_job_fit, generate_cover_letter, and rewrite_cv_for_job.

Import a resume (also called CV — terms interchangeable) from plain text. Parses the text and persists it as a new CV profile in the user's candidate-profiles library. Returns the new cv_profile_id for use with analyze_job_fit, generate_cover_letter, and rewrite_cv_for_job.

List the user's CV profiles (CV and resume are interchangeable terms — this tool covers both) as a summary view (id, profile_name, display_name, headline, created_at). For full parsed_content (skills, experience, education), use the returned id with analyze_job_fit, generate_cover_letter, or rewrite_cv_for_job. Backend ordering is created_at DESC; if you omit cv_profile_id on downstream tools, the most-recent (index [0]) is used.

List the user's AI-rewritten CVs (also called resumes — terms interchangeable in this MCP) generated by rewrite_cv_for_job. Returns a summary view (id, name, original_cv_profile_id, jd_id, focus_type, template, created_at). Backend ordering is created_at DESC; index [0] is the most recent. To download a previously-generated CV, call get_cv_download_url(cv_id, format) — note format is fixed at the time of generation: if you generated a docx and now want PDF, you must call rewrite_cv_for_job(..., export_format='pdf') again. There is no cross-format on-demand re-render in v0.4.0.

List your Jobs Module library as a summary view (id, title, company, location, brief summary). For full parsed_content / classification codes / salary bands, call get_job with the returned id. For active-pipeline rows with status, call list_tracked_jobs.

List jobs the user is actively pursuing (Job Tracker pipeline — saved/applied/interviewing/offer/not_offered/accepted). Returns paginated tracker rows with status metadata. For the full library of imported JDs, use list_jobs.

Generate a CV (also called a resume — terms interchangeable in this MCP) tailored to a specific job description. Optionally persists and returns a structured export descriptor: data.export.url (long signed download URL), data.export.short_url (chat-client-friendly short URL — new in v0.3.0), expires_at, content_type, size_bytes, filename. Set export_format to "docx" (default) or "pdf" — both formats route through the same /d/[code] short-URL flow. Use get_cv_download_url to mint a fresh short URL after the original expires; the re-mint preserves the original format (no cross-format re-render). Note: MCP-generated PDFs always render with contentType='standard' and length='3_page' regardless of what variation you previewed in the browser UI — the MCP layer does not expose those knobs in v0.4.0. To get a different variation, change the variation in the browser UI and use its Export PDF button. PDFs render with all sections visible; if you toggled sections off in the browser UI, that state does not propagate to MCP PDFs.

Generate a signup URL for a new user. Includes MCP attribution and client tracking parameters. No authentication required.

Add a job to the user's Job Tracker pipeline (the same tracker shown at /job-tracker, listed by list_tracked_jobs). Use after import_job_from_url / rewrite_cv_for_job to record that the user is pursuing a role. title and company are required; status defaults to "Saved" (valid: Saved, Applied, Interviewing, Offer, Not Offered, Accepted). Pass jd_id to link the imported JD and linkedCvId to link a generated CV/résumé. Returns the new tracker_id.