BrightHire

Get AI-generated Q&A notes for one or more interviews (up to 20). Returns condensed question-answer pairs automatically extracted from the interview transcript. Each note contains the interviewer's question and the candidate's answer, with timestamps. Each note's `start_time` is seconds from recording start — use it as the `?t=` query parameter on `https://app.brighthire.ai/interview/<call_id>?t=<seconds>` deep links (same meaning as transcript segment `start_time`). Each returned call also includes the associated `scheduled_interview_id` (when present) so guide-coverage workflows can join AI notes to `get_interview_guide` results without re-querying `search`. Each returned call also includes `call_timestamp_display`, a rendered weekday/date/time for the interview; quote it verbatim rather than deriving a weekday or relative day from `call_timestamp`. Use this instead of `get_transcripts` when you want to understand what was discussed without consuming a large context window. AI notes are much smaller than full transcripts but capture the key content. Notes are generated with redacted transcript content left out. Each returned call includes `notes_status`: `COMPLETED`, `IN_PROGRESS` (notes are still generating), `FAILED`, or null (notes were never generated for this call). Use it to explain empty `notes`; `COMPLETED` with empty `notes` means no notes were produced or none matched `keyword`. Accepts up to 20 call_ids at once, so you can batch-fetch notes for multiple interviews from a single search result page. The call_ids can be obtained from search results. Use this tool when: Analyzing interview quality or candidate responses; Comparing answers across multiple interviews for the same position; Summarizing what was covered in an interview; Evaluating interviewer question quality. Each note's `question` field is AI-generated from the transcript and may paraphrase what was said — use judgment to semantically match notes to topics (e.g. interview guide cues) rather than relying on exact keyword overlap alone.

Get the most recent completed candidate summary notes for a candidate. Returns ordered summary notes with topic, subtopic, text, and citations (each with the call_id and time in seconds locating the cited interview moment). Citations to interviews the requesting user cannot view are omitted, so an empty citations list does not mean the note had no citations.

Get the calling user's most recent BrightHire AI coaching report as raw markdown. Coaching reports summarize trends and feedback across the user's recent interviews. Takes no arguments — the user is inferred from the MCP session, and only their own report is returned. Returns the report as a single markdown text block. If the user has no coaching report yet, the tool returns an error indicating the report was not found.

Get a single custom AI summary template by `template_id`. Returns the full template definition: name, `custom_prompt`, ordered `sections` (title + prompt bullets), visibility, default settings, and meeting type. Private templates owned by another user are not accessible — the tool returns an error in that case. Obtain `template_id` values from `list_custom_templates`.

Get interview prep context for a scheduled interview. Returns scheduled start time (offset-aware ISO string in the requesting user's timezone, plus `scheduled_start_display`, the same moment already rendered with weekday, date, time and timezone — quote it verbatim rather than deriving any of those yourself), candidate and position metadata, action URLs, interview guide, and candidate summary (whose notes carry citations locating the cited interview moment; citations to interviews the requesting user cannot view are omitted).

Get interview guides for one or more scheduled interviews (up to 20). Returns a list of guides, each with the guide name and an ordered list of questions/cues to help the interviewer prepare. Accepts up to 20 scheduled_interview_ids at once so you can batch-fetch guides for a full page of search results. For large batches, pass `fields: ['cue']` to omit long description text and shrink the response dramatically.

Get a post-interview summary for a single interview by interview_id. Returns candidate name, position title, user-created interviewer notes (text notes, cues, comments), emoji reactions (thumbs up/down, star, flag) with counts, and action links (interview_url, share_url, ats_scorecard_url). Use this to build a recap of what the interviewer captured during the call. For AI-generated Q&A notes derived from the transcript, use get_call_ai_notes instead.

Get one or more job descriptions. Pass `position_id` for a single role or `position_ids` (1-50 UUIDs) to batch-fetch JDs for multiple positions in one call. Use this to understand role requirements, key skills, qualifications, and responsibilities. position_ids can be found in search results (each interview is associated with a position). This is especially useful for: Understanding what skills and competencies to evaluate when analyzing interviews; Comparing candidate responses against the actual job requirements. On success returns {job_descriptions: [{position_id, title, ats_job_posting_id, ats_job_posting_updated_at, description}]}. Positions you can't view, that don't exist, or that have no ATS JD attached are silently omitted from the response — diff the input set against the returned position_ids if you need to know which were dropped. If every requested position would be dropped, the tool returns an error result instead of an empty payload, so you don't need to special-case an empty list — just stop and surface the error to the user.

Get Knowledge Base documents the authenticated user can view. Returns org-level knowledge assets plus, when `position_id` is provided, assets scoped to that position. Each result includes id, name, description, category, and (when available) a bounded excerpt of the document's extracted text. When `text_truncated` is true, call `read_knowledge_document` with that asset's `id` (and the same `position_id`) to read the full text. Requires the organization to have Knowledge Base enabled; otherwise returns an error. Access control is enforced server-side — only assets the caller is permitted to view (by organization membership and folder/asset access grants) are returned.

Get upcoming scheduled interviews. Returns metadata for each scheduled interview including candidate name, position title, interviewers, and scheduled start/end times. Each item carries `scheduled_start` / `scheduled_end` as offset-aware ISO strings in the requesting user's timezone and `scheduled_start_display` / `scheduled_end_display` as the same moments already rendered with weekday, date, time and timezone — quote the display values verbatim and never derive a weekday or local time from the ISO strings yourself. By default returns interviews scheduled from now through the next 7 days. Use `from_date` and `to_date` to customize the time range (accept either a Unix timestamp in seconds OR an ISO 8601 date/datetime string; naive dates and datetimes are interpreted in your timezone). Set `current_user_only: true` (default) to only return interviews where you are a scheduled interviewer. Set `current_user_only: false` to see all scheduled interviews you have access to in your organization. The `scheduled_interview_id` in results can be used with `get_interview_context` to get full interview prep details including the interview guide and candidate summary.