acknowledge_legal_documents
ChatGPTRecord the candidate's acceptance of the job's mandatory legal documents, clearing the legal-doc gate on submit_application. Call ONLY after the candidate has been shown each document link (from get_job_details/start_application) and has explicitly agreed — never on their behalf. Each call ADDS to the accepted list: documents accepted earlier stay accepted, so a later call never drops a consent. IDs must come from get_job_details → legal_documents[]; an unknown ID is refused with the list of valid IDs.
answer_screening
ChatGPTAnswer the CURRENT screening question for the pending application (questions come one at a time; each response shows the next). Repeat until all required questions are answered, then submit_application. Answer format by type: text/textarea/numeric/numeric_range/date/date_range → the value directly; single_choice → the option number or exact label/value; multiple_choice → comma-separated option numbers/values.
check_login_status
ChatGPTCheck the current authentication status of the user.
get_cv_json_schema
ChatGPTINTERNAL plumbing — do NOT narrate. Returns the canonical CV JSON shape; use it between get_cv_text and upload_cv_data to parse the extracted text. Frame as "reading your CV", never "fetching a schema".
get_cv_text
ChatGPTINTERNAL plumbing — do NOT narrate "fetching from S3"/"extracting PDF"; frame as "reading your CV". Returns extracted plain text from a stored CV (pdf/doc/docx/rtf/txt/md). Sequence: select_file_for_application({ kind: "cv" }) → this → parse to get_cv_json_schema shape → upload_cv_data.
get_job_details
ChatGPTRetrieve the full job profile — description, application requirements, screening questions, employer info, and any legal documents. CALL before start_application, or when the user wants to read a job before applying.
get_upload_link
ChatGPTGet a browser URL the user visits to upload a new file of the given kind (the page handles sign-in and the S3 upload). After they confirm, call list_my_files({ kind }) to see the new file. Accepted: pdf, doc, docx; max 5 MB; max 5 files per kind per candidate.
list_my_files
ChatGPTList the candidate's confirmed files of the given kind (up to 5): file id, label, upload time, size, and — for CVs — whether structured JSON data is attached. Call before start_application when the job needs a CV or cover letter; if the list is empty, call get_upload_link({ kind }) and have the user upload via the browser — never submit without a required CV.
logout
ChatGPTLog out the current user and clear the session.
my_applications
ChatGPTList the candidate's applications, drafts included (employer, job title, pipeline status, whether it is a draft / withdrawn / ended, the submitted or started date, source). Call when they ask about their application history or whether they already applied to a job. No arguments.
prefill_answers
ChatGPTRecord answers to several unanswered screening questions at once, drawn from the candidate's CV. Already-answered questions and invalid values are skipped. Show the candidate what you are about to record and get their agreement first — the answers are saved on the draft, not just suggested. ChatGPT compatibility note: pass the suggestions as a JSON STRING, not a free-form object, because ChatGPT drops tool definitions whose parameters contain an object with no declared properties.
process_cv
ChatGPTRender the pending application's CV data (from the session or selected CV file) plus the unanswered screening questions, for reasoning over the CV to suggest answers. Call AFTER start_application created a draft AND the selected CV has structured data.
save_job
ChatGPTSave a job to your bookmarks for later review.
saved_jobs
ChatGPTList all saved/bookmarked jobs.
search_jobs
ChatGPTSearch Brave New Jobs for open positions; applications are built in conversation. CALL when the user wants to find work ("find me a job", "roles in {location}", "what positions are open", "jobs matching my CV", etc.). Filter by keywords, location, salary, employment type, workplace type. Requires the candidate to be signed in, like every tool on this server. After returning matches, offer start_application({ job_id }) for any role they are interested in.
select_file_for_application
ChatGPTPin a specific file (chosen from list_my_files({ kind })) to the current application; verifies ownership + confirmed upload and stores it on the session for start/submit. For kind: "cv": after this succeeds you can call start_application immediately — the attached file is enough to open the draft. Structured data is required only before submit_application (not before start): if the chosen CV has none yet, call get_cv_text, parse it into the get_cv_json_schema shape, then upload_cv_data({ cv_file_id, structured_cv }) — any time before submitting. For kind: "cover_letter": the cover letter is attached to the pending application and submitted when submit_application runs.
start_application
ChatGPTOpen a DRAFT application (NOT the final submission). Nothing reaches the employer until submit_application succeeds; do not tell the user they have applied until it does. Before calling, fetch get_job_details({ job_id }) and satisfy fieldRequirements. CV (if cv === "required"): JSON-only jobs (cv_original_file === "optional") → parse the CV shared in-chat into get_cv_json_schema shape and call upload_cv_data({ structured_cv }) (no file); file-required jobs → list_my_files({ kind: "cv" }), else get_upload_link({ kind: "cv" }) and wait, then select_file_for_application({ kind: "cv", file_id }) (and if it has no structured data, get_cv_text → parse → upload_cv_data({ cv_file_id, structured_cv })). Cover letter must be attached before/at draft creation: an uploaded file via select_file_for_application({ kind: "cover_letter" }) first, or inline via this call's cover_letter arg (pass the actual text, not intent). After the draft opens, use update_application({ cover_letter_text }). Screening questions (if any) are presented one at a time via answer_screening. After the draft: answer the screening questions, settle the cover letter, then recap and submit. On failure the response names the next tool to call.
submit_application
ChatGPTSubmit the completed application. The server re-validates every requirement and, on failure, returns the specific missing-field codes and names the exact next tool to call — follow it. REFUSED unless: every required screening question is answered (answer_screening/prefill_answers); required file fields are attached; mandatory legal docs are accepted via acknowledge_legal_documents; the parsed CV is present (file-required jobs — cv_original_file absent/"required" — also need an original CV file, not just structured data); the CV name matches the account (else resend cv_identity_confirmed: true); and you showed the candidate a full recap they confirmed (pass confirmed: true only then). [spec 013 / spec 059]
update_application
ChatGPTPatch fields on the current application. Pass ONLY the fields you want to set (any combination): expected_salary — the candidate's expected pay (annual gross by default; currency defaults to EUR, period to "year"). Required when the job has fieldRequirements.expected_salary === "required" (submit is otherwise refused). Requires an open draft (start_application first). notice_period_days — how soon the candidate could start, in days. Required when the job has fieldRequirements.notice_period === "required" (submit is otherwise refused). Pick the bucket matching what the candidate says: 0 = immediately, 7 = 1 week, 15 = 15 days, 30 = 1 month, 45 = 45 days, 60 = 2 months, 90 = 3 months, 91 = more than 3 months. Requires an open draft. cover_letter_text — the FULL inline cover-letter text (the actual letter, NOT a placeholder). Works at ANY point: attached immediately if a draft is open, else staged and attached when start_application runs. Re-calling REPLACES the prior text. To attach an uploaded FILE instead, use select_file_for_application({ kind: "cover_letter" }). allow_inference — whether the assistant may use CV data to suggest answers to screening questions. Requires an open draft.
upload_cv_data
ChatGPTINTERNAL plumbing — do NOT narrate "parsing JSON"/"uploading data"; frame as "reading your CV". Attaches the structured JSON CV (parsed into the get_cv_json_schema shape) to the application. Required before applying to a job with fieldRequirements.cv = "required". TWO modes: (a) file-backed — pass cv_file_id (from list_my_files({ kind: "cv" }), after select_file_for_application({ kind: "cv" }) → get_cv_text → parse) to attach the JSON to an uploaded CV file. (b) JSON-only (parsed-CV-only jobs) — OMIT cv_file_id: parse the CV the candidate shared in this chat and pass it as structured_cv, no file, no upload link.