automation_delete
ChatGPTRemove an automation from a posting. Requires a confirm_token from automation_delete_preview. Archived, not destroyed.
automation_delete_preview
ChatGPTREQUIRED before automation_delete. Returns a confirm_token.
automation_list_for_posting
ChatGPTShow the workflow automations set up on a posting and whether each is on. One that is on runs the user's own rules by itself and may email candidates; check this before moving candidates.
automation_list_templates
ChatGPTList the company's reusable automation templates and what each does. Read-only: automations are added to postings, built and turned on in the Lighthouse app, not here.
candidate_add_note
ChatGPTAttach an internal note to a candidate. Visible to the hiring team in Lighthouse; the candidate never sees it. Use it to record what the user decided or discussed.
candidate_get_details
ChatGPTReads one applicant's record in Lighthouse: contact details and current role, stage, tags, the Lighthouse score with its four component scores, the star rating, the skills, roles and industries the scorecard matched, years of relevant experience, screening answers, and the 10 latest team notes. Read-only.
candidate_list_for_posting
ChatGPTList the applicants on a posting that match the user's filters (stage, minimum Lighthouse score), with the score Lighthouse already calculated and their current stage, ordered by that score. Use for 'who applied', 'who scored 8 or higher', 'who is in Phone Screen'.
candidate_move_stage
ChatGPTMove ONE candidate to a different interview stage (never to Will Not Proceed: rejecting is done in Lighthouse). Requires a confirm_token from candidate_move_stage_preview and the user's explicit agreement. May trigger automations.
candidate_move_stage_preview
ChatGPTREQUIRED before candidate_move_stage. Shows who moves, from where to where, and whether any automation the user turned on will run (which may email the candidate). Returns a confirm_token.
candidate_set_rating
ChatGPTRecord the 1-5 star rating the user gives a candidate, or clear it. Only set the rating the user states; never choose one yourself. Kept separate from the Lighthouse score.
candidate_set_tags
ChatGPTReplace the tags on a candidate. Tags are internal labels for filtering in Lighthouse — the candidate never sees them. Pass the FULL list you want; omitted tags are removed.
candidate_show_side_by_side
ChatGPTReads 2-10 applicants' records for side-by-side display: Lighthouse score and its four components, star rating, stage, current title, years of relevant experience and matched skills, listed in order of Lighthouse score. Read-only.
email_get_draft_context
ChatGPTReturns the candidate's name, email, role and stage, plus a saved template if you name one, so YOU can write a specific draft for the user to review. It does not write the subject or body itself, and NOTHING IS SENT — the user sends from Lighthouse. Always tell them it has not been sent.
email_list_templates
ChatGPTList the company's saved candidate-email templates. Offer these before writing from scratch — the wording is usually already agreed, and templates carry the right tokens.
interview_get_grades
ChatGPTShow how a candidate was graded in their interviews — the questions asked, the interviewer's notes, and any scores. Use this alongside candidate_get_details when the user wants a candidate's full record.
interview_get_stage_questions
ChatGPTList the questions interviewers ask at one stage. Every candidate who reaches that stage is asked these, and interviewers record their grades against them.
interview_set_stage_questions
ChatGPTReplace the interview questions on a stage, in the order they should be asked. Previous questions are archived rather than destroyed, so grades already recorded stay readable.
interview_suggest_questions
ChatGPTDraft three interview questions for one stage, using the same AI as the app's Manage stages panel. It reads the role, the job description and the questions already on every stage, so it avoids repeats. This only SUGGESTS them — show them to the user, take their picks, then save with interview_set_stage_questions. Call again for more.
list_my_companies
ChatGPTThe companies this user is a member of, and which one is their default. Use it when the user mentions another company, to check they have access and get the exact name or id to pass as company to other tools.
posting_archive
ChatGPTArchive ONE posting (the app's Delete). Requires a confirm_token from posting_archive_preview and the user's explicit agreement.
posting_archive_preview
ChatGPTREQUIRED before posting_archive. Archiving is the app's Delete: the posting disappears from every list, report and the careers page. The record and its candidates are kept, not destroyed. Works on drafts and published postings.
posting_build_scorecard
ChatGPTBuild a DRAFT posting's scorecard from its accepted search criteria, using Lighthouse's scorecard service, and wait for it (about a minute). Only needed when posting_draft_status or posting_publish_preview asks for it: posting_criteria_accept already starts this in the background. Does nothing if the posting already has a scorecard.
posting_criteria_accept
ChatGPTSave the search criteria the user approved on a DRAFT posting, then start building the posting's scorecard from them in the background (Lighthouse's scorecard service; never written by you). Only call after the user approves the list from posting_criteria_draft. Can be done once per posting.
posting_criteria_draft
ChatGPTHave Lighthouse draft the search criteria for a DRAFT posting: who the user is looking for. Lighthouse builds the posting's scorecard from these once the user accepts them. First ask the user who they are looking for (must-haves, nice-to-haves, deal-breakers) and pass their answer in looking_for, or set use_job_description_only if they want it drafted from the job description alone. Takes 10-40 seconds. Nothing is saved until posting_criteria_accept.
posting_draft_set
ChatGPTSet one or more fields on a draft posting. Pass only the fields you have — omitted ones are left alone. Returns the updated draft with anything still missing or invalid.
posting_draft_start
ChatGPTBegin a new job posting for the user's company. Creates a DRAFT that neither candidates nor the Lighthouse app shows. Returns a draft_id plus exactly what is still needed — ask the user for those, then call posting_draft_set. posting_publish then creates it as Open (live) or Paused (in Lighthouse, hidden from candidates).
posting_draft_status
ChatGPTShow a draft's current values and what is still needed before it can be published.
posting_generate_description
ChatGPTWrite a job description for a draft posting, using the same AI the app uses. This only DRAFTS it — show the result to the user, take their edits, then save it with posting_draft_set. Ask for extra context first; it materially improves the output.
posting_list
ChatGPTList the company's job postings. Returns published postings by default; pass include_drafts to see drafts the user has started but not published.
posting_publish
ChatGPTTurn a draft into a real posting, Open (visible to candidates immediately) or Paused (in Lighthouse, hidden from candidates). Requires a confirm_token from posting_publish_preview and the user's explicit agreement.
posting_publish_preview
ChatGPTREQUIRED before posting_publish. Returns a plain-English summary of exactly what will be created plus a confirm_token. Show the summary to the user and get their explicit agreement, then pass the token to posting_publish. The token lasts 5 minutes and works once. Drafts are invisible in Lighthouse, so if the user wants the role set up but not live yet, publish it with status 'Paused'.
posting_set_status
ChatGPTChange a published posting's status. Requires a confirm_token from posting_set_status_preview and the user's explicit agreement.
posting_set_status_preview
ChatGPTREQUIRED before posting_set_status. Paused and Closed hide the job from the careers page and job boards; Open shows it. Returns a summary and a confirm_token.
report_pipeline_snapshot
ChatGPTWhere every active candidate currently sits, how long they have been waiting, and how many are stalled. Use this for 'how are we doing', 'what needs attention', 'who is stuck'. Covers the whole company, or one posting if you pass posting_id.
report_posting_funnel
ChatGPTHow candidates on one posting moved through the pipeline: how many applied, reached each stage category, were hired, and did not proceed. Use for 'how is this role performing'.
report_roles_across_companies
ChatGPTOpen postings in ALL of the user's companies, not just the connected one, with how long each has been open in business days and how many candidates are waiting. Use this for 'which roles are going stale', 'anything open more than 10 business days', or a recurring check across a portfolio of clients. There is no company argument — it always covers exactly the companies this user is a member of. Read-only. Other tools work in one company per call (the default, or the one passed as company).
report_time_to_stage
ChatGPTTime-to-stage hiring metrics such as time to fill, time to hire, time to submit or time to interview. For each role: the date it opened, how many business days it has been open, and whether any candidate has reached a given stage yet (for example Submitted to Hiring Manager, Phone Screen, or Hired). If one has, it gives the date the first candidate entered that stage and the business days from opening to then. Optionally compares each role with a target such as 10 business days. Use it for questions like 'how long has this role been open and has anyone reached the hiring manager yet', 'what is our time to hire this quarter', or 'which roles have no one in interview after 2 weeks'. Covers the active company by default, or several of the user's own companies. Business days are weekdays; public holidays are not excluded. Read-only.
role_template_list
ChatGPTThe company's role templates: each one's role, defaults and which scorecard, stage workflow, automation and questionnaire templates it links. New postings can start from these in Lighthouse.
scorecard_generate
ChatGPTDraft a scorecard (requirements and weights) from a posting's job description and the user's own must-haves, using Lighthouse's scorecard generator. This only DRAFTS it: nothing is saved and no score changes until scorecard_save. Show it to the user to edit first.
scorecard_get_active
ChatGPTShow a posting's scorecard: the requirements and weights the user set, which Lighthouse uses to calculate each applicant's 0-10 résumé match score. Four weighted components, with the requirements under each. Returns nothing if the posting has no scorecard yet, in which case Lighthouse calculates no scores for it.
scorecard_list_templates
ChatGPTList the company's saved scorecard templates (requirements and weights). Offer these before drafting a new one; teams usually reuse their own.
scorecard_save
ChatGPTSaves the scorecard shown in scorecard_save_preview as a new version of the posting's scorecard and switches the posting to it (earlier versions are kept). Lighthouse then recalculates the match score of every applicant on the posting in the background, which takes a few minutes. No stage changes. Requires the confirm_token from scorecard_save_preview; only call it after the user agrees.
scorecard_save_preview
ChatGPTREQUIRED before scorecard_save. Shows what saving this scorecard will do: it becomes a new version of the posting's scorecard (earlier versions are kept), and Lighthouse recalculates the match score of every applicant on the posting against it. No stage changes. Writes nothing; returns a confirm_token.
screening_get_answers
ChatGPTShow what a candidate answered when they applied, including any answer flagged by a knockout question.
screening_get_posting_questions
ChatGPTList the questions applicants answer when applying to a posting, in order. A knockout question flags an answer that does not match the answer the user requires (e.g. work authorization). Lighthouse shows the flag with their answers.
screening_list_questionnaires
ChatGPTList the company's reusable question sets. Offer these before writing questions from scratch — teams usually have a standard set they want reused.
screening_set_posting_questions
ChatGPTReplace the screening questions on a DRAFT posting. Pass them in the order applicants should see them. Existing questions not in your list are archived, never hard-deleted, so any answers already collected survive.
search
ChatGPTFind postings or candidates by name across this company. Use it to turn something the user said ('the finance analyst role', 'Priya') into an id for other tools.
template_archive
ChatGPTArchive ONE template (Template Management's Delete). Requires a confirm_token from template_archive_preview and the user's explicit agreement.
template_archive_preview
ChatGPTREQUIRED before template_archive. Archiving is Template Management's Delete: the template disappears from every picker, but it is kept, and postings already made from it are unaffected. Warns when role templates still link it.
template_from_posting
ChatGPTCreate the templates previewed by template_from_posting_preview. Pass the same arguments plus the confirm_token, after the user's explicit agreement. A part that fails is reported, and the rest still saves.
template_from_posting_preview
ChatGPTREQUIRED before template_from_posting. Lighthouse's 'Create Template' on a posting: copies its active scorecard, stages, automations and screening questions into new templates, then a role template that links them, with the posting's role details as defaults. The posting itself is not changed.
template_get
ChatGPTRead one template of any kind in the shape template_save_preview takes, so it can be changed and saved back with its template_id. Automations come back as a summary plus their stage names.
template_save
ChatGPTSaves the template shown in template_save_preview: creates it, or replaces the one with that template_id. Pass the same arguments plus the confirm_token; only call it after the user agrees. Replacing an email template changes what automations that send it send from then on.
template_save_preview
ChatGPTREQUIRED before template_save. Shows what saving a template (email, scorecard, stage workflow, screening questionnaire or role) would create, or what it would replace when template_id is given. Writes nothing; returns a confirm_token. Templates are company-wide. Scorecard, workflow and questionnaire templates are copied onto new postings, so saving one changes no existing posting or candidate. Email templates are used as they are by any automation that sends them, so replacing one changes what those automations send from then on.
whoami
ChatGPTIdentify the signed-in Lighthouse user and the company this connection acts on. Call this FIRST in any session so you know whose data you are working with. This is the default company; pass company to any tool to work in another one the user belongs to.
workflow_get_stages
ChatGPTList the interview stages candidates move through for a posting, in funnel order.
workflow_list_templates
ChatGPTList the company's saved stage workflows, plus the built-in default. Offer these to the user rather than inventing a funnel from scratch.
workflow_set_stages
ChatGPTReplace the interview stages on a DRAFT posting. Pass stages in funnel order — the order of the array is the order candidates move through. 'Hired' and 'Will Not Proceed' are appended automatically; do not include them. Pass no stages to use the standard default.