add_companies_to_company_list
ChatGPTAdds companies to an existing user company list. AGENTS: companyIds already present in the list are ignored server-side, so it is safe to re-send. The updated list (with new companyCount) is returned.
add_persons_to_person_list
ChatGPTAdds private contacts (house-list persons) to an existing user person list. AGENTS: privateContactIds already present in the list are ignored server-side, so it is safe to re-send. The updated list (with new personCount) is returned.
create_company_list
ChatGPTCreates a new user company list, optionally seeded with a set of companies.
create_person_list
ChatGPTCreates a new user person list, optionally seeded with a set of private contacts (house-list persons).
create_phone_calls
ChatGPTRegisters phone calls that were already made outside Sales Marker (for example a client's own Excel / Google Sheets call log) as approach history. Each entry is stored as a phone call in the given call list and its outcome is recorded as a project action on the company's project for the list's product, exactly as if the call had been logged from the phone call list UI. AGENTS: This is a WRITE tool that creates approach history — call it only on an explicit user request to import or register calls. phoneCallListId must reference an existing call list that has a product (use list_phone_call_lists to find one). Every entry needs companyId (法人番号), phoneNumber and result; result is the call outcome and becomes the project phase name, e.g. "アポ獲得", "受付NG", "担当者NG", "不通", "コール済み". Omit calledAt to record the call as of now. Re-sending the same rows is safe: an entry whose phone call already has an action with the same calledAt and result is skipped and counted in skippedCount, so a sheet can be re-synced without duplicating history. An entry that is newer than the call's latest recorded action overwrites that call's stored result, which is why this tool is annotated as destructive. At most 100 entries per call; the whole batch is rejected if any entry is invalid.
create_sequence
ChatGPTCreate a sequence for the caller's subscription. The new sequence is in the CREATING state and approaches nobody until StartSequence is called. The member source depends on listType: COMPANY_LIST takes userCompanyListId; SIGNAL_LIST and PERSON_LIST require at least one trigger. With templateNodes empty (the "customize" template in the UI) the sequence gets only its fixed end steps, and approach steps are added one at a time with AddNode. A non-empty templateNodes pre-fills approach steps with default content. AGENTS: Always create with templateNodes empty (the customize template) and add approach steps afterwards with AddNode; never send a pre-filled template. Pick productId and targetId from get_products and get_all_targets. memberAddMode defaults to MEMBER_ADD_MODE_PERIODIC and memberType to MEMBER_TYPE_COMPANY when omitted. The response only carries the new sequenceCommonId; use get_sequence to read the sequence back.
delete_company_list
ChatGPTPermanently deletes a user company list and its company memberships. AGENTS: this is irreversible; confirm with the user before calling. Returns FAILED_PRECONDITION when the list is still used as an existing-customer list of a product or by a workflow, and PERMISSION_DENIED when the list belongs to another user and is not editable.
delete_sequence
ChatGPTDelete a sequence. Its members and engine state are removed, running advertisements are paused and its email, form and SNS lists are disabled. AGENTS: Irreversible through this API. Confirm with the user first, and prefer pause_sequence when they only want approaches to stop.
duplicate_sequence
ChatGPTCopy a sequence into a new CREATING sequence owned by the caller. Settings, triggers, exclusions, nodes and their approach lists are copied; members are not.
get_all_targets
ChatGPTReturns all intent segments for the current user's subscription. AGENTS: Takes no required arguments; targets are scoped to the caller's subscription. A target represents an intent segment tied to a product.
get_company_approach_results
ChatGPTGets the project actions (approach results) toward one target company. AGENTS: Use when asked what happened with a specific company; for a product's outreach across companies use get_product_approach_results instead. Omit endActDate for a single day (it is exclusive).
get_company_labels
ChatGPTGetCompanyLabels is exposed as OpenAPI GET /v2/search/company-labels. It returns the enabled company labels (tags) that can be used as company filters, each with its id, name, and type. Labels are shared Sales Marker master data, the same set shown in Sales Marker Web; they are not specific to a subscription. AGENTS: Call this before SearchCompanyV2, SearchPersonV2, or SearchDepartmentV2 whenever the user asks for activity tags (活動タグ), business content / business field tags (分野タグ), or commerce type tags (業態タグ). Match the user's wording to the returned label names and pass the IDs of the matched labels to companyFilter.activityTagIds, businessFieldTagIds, or commerceTypeTagIds according to each label's type. Narrow the result with types and nameQuery. Never invent label IDs; ask the user when no clear match exists. Do not use get_all_targets for these filters.
get_company_list
ChatGPTFetches a single user company list by id, including its name, memo and company count.
get_current_user
ChatGPTGet current user UserID is inferred from the authenticated userContext making the request. AGENTS: Use this to identify the caller before other lookups. Takes no arguments; the user is resolved from the request context.
get_domain_list
ChatGPTGetDomainList returns the subscription's watchlist domains, each with its id, domainId, domain, name, category, and status.
get_engage
ChatGPTReturns the engage settings of the caller's subscription (subscription_engage row). The subscription is resolved from the authenticated user in the gRPC context, not from the request. Returns NOT_FOUND when the subscription has no engage row. AGENTS: Takes no arguments; pass {}. NOT_FOUND means the subscription has no Engage configuration yet.
get_industries
ChatGPTGetIndustries returns industries grouped by main category with sub-industries. Use this to resolve Japanese industry names to IDs for companyFilter.industryIds and excludeIndustryIds. industries[].id is a main category and covers all of its sub-categories when searched; industries[].children[].id is a single sub-category.
get_intent_score_versions
ChatGPTGetIntentScoreVersions returns available intent score versions for the subscription. AGENTS: Returns the intent-score versions active for the current subscription. Takes no arguments; the subscription is resolved from context.
get_intent_stats_for_graph
ChatGPTGetIntentStatsForGraph returns daily keyword-intent activity for the given companies, with one row per company, target segment, keyword, and JST day. Each row has the number of devices that showed intent for the keyword on that day. This is the data behind the Sales Signal (セールスシグナル) graph on the company details screen. Only the target segments the authenticated user can access are searched, the same targets that get_all_targets returns; this includes both active and past targets. AGENTS: Use to show how a company's keyword intent changed over time, per target segment and keyword. companyIds are corporate company IDs. If the user gives a company name, MUST call search_company_v2 first and use its companyId; if the user gives a corporate company ID, use it directly. Before passing segmentIds or keywordIds, MUST call get_all_targets with read_mask.paths containing "id", "name", "keywords.id", and "keywords.name"; use target.id as segmentIds and keyword.id as keywordIds. Never invent IDs. When both are given, every keywordId must belong to one of the segmentIds. Empty segmentIds means every target the user can access; empty keywordIds means every keyword of the selected segments. For the whole graph, omit both and do not call get_all_targets: each row already has segmentName and keyword for labels. Leave version empty to use the subscription's current intent-score version, as the Sales Marker screen does; pass a version from get_intent_score_versions only when the user asks for a specific one. fromDate/toDate are inclusive YYYY-MM-DD (JST) and default to the 30 days through today; the screen offers 7, 30, and 90 days. Days without activity are omitted; treat a missing day as zero devices. Rows are ordered by companyId, segmentId, keywordId, then date ascending. Use limit/offset and hasMore to page; keep every filter unchanged between pages. Before drawing a chart, fetch every page: set limit to 1000 and repeat with offset increased by limit while hasMore is true. To reproduce the Sales Signal graph, group rows by companyId and segmentId and draw one chart per company and segment with one series per keywordId. Use date on the X-axis and totalNumDevices on the Y-axis, and add a zero-valued point for every missing date in the requested range. Keywords with no activity in the period return no rows and are not shown. Do not combine different companies or keywords from different segments. Results include both active and past targets. The Sales Marker screen shows active targets (at least one keyword active, stop requested, or application in progress) by default and past targets in a separate tab; say that both are included when presenting the graph.
get_job_listing
ChatGPTGetJobListing returns one job listing (求人) with its full job description and dispatch (haken) predictions. The response contains the same summary fields as search_job_listings plus the full description and the predicted dispatch (haken) companies. ACCESS: Only subscriptions with the job-listing feature enabled (subscription.canUseJobListingTable) can call this RPC; others receive PERMISSION_DENIED. AGENTS: WHEN TO CALL: Call GetJobListing (MCP get_job_listing, OpenAPI GET /v2/search/job-listing/{id}) when the user wants the full description or details of a job from search_job_listings results, e.g. "show the description", "詳細を見せて", or selecting/clicking a job. - id: MUST be a jobListings[].id returned by search_job_listings (the job-listing ID). Never pass a companyId and never invent IDs. If the user refers to a job by title or position in the list, map it to the id from the latest search results; ask the user when it is ambiguous. Call it once per job the user wants to read, not for every search result. Show the description as returned; summarize it only when the user asks. ERRORS: On NOT_FOUND, tell the user the job listing no longer exists. On PERMISSION_DENIED, tell the user their plan does not include job-listing search and do not retry.
get_lead_actions
ChatGPTLists the marketing lead actions recorded against one private contact. Each row carries the date it happened, an optional free-text memo, the user who took it, and ids identifying where it came from (inquiry, material, pop-up, scheduler or sequence) together with its type and result ids. AGENTS: privateContactId is required; this tool does not search across leads or by date range. typeId / typeResultId are lead_type / lead_type_result ids with no name lookup tool yet, so report them as ids. isNextAction=true rows are planned follow-ups, not completed actions. Returns an empty list when the lead has no actions; PERMISSION_DENIED means the caller has no subscription.
get_meeting_details
ChatGPTGet full meeting details for a single meeting, including a presigned transcriptURL, AI notes, and meeting attributes. AGENTS: Call after search_meetings to fetch a meeting's transcript; download the presigned transcriptURL to read its content.
get_member_history_for_node
ChatGPTList the members that have a history record at one node of a sequence. Optionally restrict to a history type (APPROACH, NG, APPOINTMENT or UNQUALIFIED) and to the history name shown in the UI.
get_note
ChatGPTThis endpoint retrieves the note for a supported object (e.g. meetings, phonecalls) AGENTS: Call with a certain interaction type (online meeting, offline meeting, etc.) and the id of the object to get the summarized note. Typically, call after search_meetings with the meeting id and meeting type of a certain item in the search response.
get_person_list
ChatGPTFetches a single user person list by id, including its name, memo and person count.
get_phone_call
ChatGPTFetches a single phone call entry, including its target and call result.
get_phone_call_list
ChatGPTFetches a single phone call list (a company-list calling campaign).
get_prefecture_list
ChatGPTGetPrefectureList is exposed as OpenAPI GET /v2/search/prefectures. Set includeCities=true to include municipality IDs for companyFilter.cityIds. When resolving a requested search location, present only the relevant matching options with their parent prefecture and ask the user to confirm the intended location.
get_product_approach_results
ChatGPTGets the project actions (approach results) for one product across all target companies. AGENTS: Use when asked how a product's outreach is going; for a single target company use get_company_approach_results instead. Pass userId to narrow to one sales rep. Omit endActDate for a single day (it is exclusive).
get_products
ChatGPTReturns all products for the current user's subscription. AGENTS: Takes no required arguments; products are scoped to the caller's subscription. Use this to discover available products before calling get_all_targets.
get_project_action
ChatGPTFetches a single project action by id, including the display names of the related user, project, company, project type and project phase.
get_properties
ChatGPTThis endpoint retrieves the properties for a supported object (e.g. meetings, phonecalls) AGENTS: Call with a certain interaction type (online meeting, offline meeting, etc.) and the id of the object to get the properties extracted from the transcription. Typically, call after search_meetings with the meeting id and meeting type of a certain item in the search response.
get_ranked_sequences
ChatGPTRank sequences by appointment rate and return the top topN. Appointment rate is appointments divided by members. Sequences with no members or no appointments are left out. AGENTS: topN must be greater than 0 and rankBy must be APPOINTMENT_RATE. The response has only sequence IDs and ranks; resolve names with get_sequence or get_sequences.
get_send_email
ChatGPTFetches a single sent/composed email, including its status and open counts.
get_send_email_list
ChatGPTFetches a single batch email request (a company-list email send).
get_send_email_template
ChatGPTFetches a single reusable email template.
get_send_form
ChatGPTFetches a single contact-form submission, including its send status.
get_send_form_list
ChatGPTFetches a single batch contact-form request (a company-list form send).
get_send_form_template
ChatGPTFetches a single reusable contact-form template.
get_sequence
ChatGPTGet one sequence by sequence_common ID. Returns its settings, triggers, exclusions, member statistics, node summary and engine instance details.
get_sequence_history
ChatGPTList the change log of a sequence. Each entry says who created, edited, started, paused, deleted, cancelled or restored the sequence, or when the batch added members, and when.
get_sequence_members
ChatGPTList the members of a sequence. Members are companies, or contacts for a person-based sequence. Each member carries the node it is currently at, its status and the nodes it has visited. nodeId restricts the result to the members at one node.
get_sequence_overviews
ChatGPTList sequences filtered by status, name, creator or member kind. Pagination is by page number, and the response counts sequences per status. Entries have the same shape as GetSequences.
get_sequences
ChatGPTList the sequences of the caller's subscription, most recently updated first. Pagination is limit/offset. Each entry carries the status and member statistics shown in the sequence list. filter restricts the result to sequences built from one saved company list. AGENTS: Use this for "show my sequences". Use get_sequence_overviews when the user wants to filter by status, name or creator.
get_subscription_ai_cost
ChatGPTEstimate what a subscription cost Sales Marker in meeting recording and AI usage over a time window, in USD. Returns one line item per cost source plus totals. Internal Sales Marker staff only; any other caller gets PERMISSION_DENIED. AGENTS: Internal staff only. subscriptionId is required. Omit endTime for now, and startTime for 30 days before endTime. Recording hours are the scheduled meeting window, not the real recording length, and offline-upload hours are what the user entered. AI rows cover only prompt-service calls logged by sales-bolt-server, so aiCostUsd is a lower bound. Labels are internal feature names. A model with no known rate is listed with costUsd 0. A subscriptionId that does not exist is not an error: it returns zero usage, the same as a subscription with no activity in the window, so check the id before reading zeros as "no usage".
get_subscription_credit
ChatGPTGet the subscription's available search credit (free, paid, and hold amounts).
get_talk_script
ChatGPTFetches a single talk script (the phone-call template).
get_user
ChatGPTGet a user by ID
get_user_company_lists
ChatGPTGetUserCompanyLists returns the user's saved company lists, each with its id, name, memo, and company count.
get_visitor_domains_and_webpages
ChatGPTGetVisitorDomainsAndWebpages returns the visitor-tracked domains, each with its hostname and the list of visited webpages.
get_visitor_intent_stats_by_path
ChatGPTGetVisitorIntentStatsByPath returns Visitor Intent page-view totals for the given companies, with one row per company, domain, and visited page path. Each row has the page views, average stay duration, and the most recent visit date for the requested period. Only the configured webpages of the authenticated user's subscription domains with Visitor Intent enabled are searched, and page views longer than 24 hours are excluded. AGENTS: Use to answer which pages a company viewed on the customer's websites, how often, and how recently. companyIds are corporate company IDs. If the user gives a company name, MUST call search_company_v2 first and use its companyId; if the user gives a corporate company ID, use it directly. Before passing domains or paths, MUST call get_visitor_domains_and_webpages: use its domain values for domains and its webpage path values for paths. Never invent domains or paths. A path matches itself and every page below it ("/function" also matches "/function/sales-signal"). When paths is empty, every authorized configured webpage is included. fromDate/toDate are inclusive YYYY-MM-DD (JST) and default to the 30 days through today. Rows are ordered by companyId, then pageViews descending. Use limit/offset and hasMore to page. Before rendering the path totals, fetch every page: set limit to 500 and repeat with offset increased by limit while hasMore is true. Only visited configured webpages are returned; an empty result means no visits to them, or that no webpages are configured. To reproduce the Sales Marker Visitor Intent view, call get_visitor_intent_stats_for_graph and get_visitor_intent_stats_by_path with identical company, domain, path, and date filters. Draw one chart per company and domain with one series per path, use zero for missing days, and render that company and domain's path totals below its graph. Do not combine different companies or domains. Convert averageDurationSeconds to a human-readable minutes/seconds value.
get_visitor_intent_stats_for_graph
ChatGPTGetVisitorIntentStatsForGraph returns daily Visitor Intent page views for the given companies, with one row per company, domain, visited page path, and JST day. Each row has the page views and average stay duration for that day. Only the configured webpages of the authenticated user's subscription domains with Visitor Intent enabled are searched, and page views longer than 24 hours are excluded. AGENTS: Use to show how a company's website visits changed over time, for example a trend per page or per day. For totals per page, use get_visitor_intent_stats_by_path instead. companyIds are corporate company IDs. If the user gives a company name, MUST call search_company_v2 first and use its companyId; if the user gives a corporate company ID, use it directly. Before passing domains or paths, MUST call get_visitor_domains_and_webpages: use its domain values for domains and its webpage path values for paths. Never invent domains or paths. When paths is empty, every authorized configured webpage is included. fromDate and toDate behave as in get_visitor_intent_stats_by_path. Days without visits are omitted; treat a missing day as zero page views. Rows are ordered by companyId, domain, path, then date ascending. Use limit/offset and hasMore to page; keep every filter unchanged between pages. Before drawing a chart, fetch every page: set limit to 1000 and repeat with offset increased by limit while hasMore is true. To reproduce the Sales Marker Visitor Intent view, call get_visitor_intent_stats_for_graph and get_visitor_intent_stats_by_path with identical company, domain, path, and date filters. Draw one chart per company and domain with one series per path, use zero for missing days, and render that company and domain's path totals below its graph. Do not combine different companies or domains. Convert averageDurationSeconds to a human-readable minutes/seconds value.
list_company_lists
ChatGPTLists the user company lists owned by the acting subscription. Use when a user asks to see, browse or pick one of their company lists.
list_intent_email_opt_outs
ChatGPTList the intent-email recipients who opted out of emails sent by the authenticated user's subscription. Results are ordered newest first. AGENTS: Use when the user asks which recipients have opted out or wants to inspect exclusions while preparing an intent-email recipient list. Intent-email sending already applies these exclusions server-side, so do not call this solely to enforce them. Paginate with limit and offset until all requested records are retrieved.
list_material_daily_stats
ChatGPTReturns per-material daily PV / CTA / CV counts (JST days) for building performance-over-time exports. AGENTS: Dates are YYYY-MM-DD in JST; both dates omitted = last 30 days. Omit materialIds to cover every material. Days without activity are omitted, so treat missing days as zero.
list_material_summaries
ChatGPTLists the subscription's engage materials with the same columns as the materials list screen (title, product, visibility, last update) plus lifetime PV / CTA / CV totals. AGENTS: Use to enumerate materials and compare their overall performance, or to resolve material ids/titles before calling list_material_daily_stats.
list_person_lists
ChatGPTLists the user person lists owned by the acting subscription. Use when a user asks to see, browse or pick one of their person lists.
list_phone_call_lists
ChatGPTLists phone call lists for the acting subscription. AGENTS: Results are paginated; when pageNumber/pageSize are omitted, server-side defaults apply. Use totalCount to decide whether to fetch more pages.
list_send_email_lists
ChatGPTLists batch email requests for the acting subscription. AGENTS: Results are paginated; when pageNumber/pageSize are omitted, server-side defaults apply. Use totalCount to decide whether to fetch more pages.
list_send_email_templates
ChatGPTLists reusable email templates for the acting subscription. AGENTS: Results are paginated; when pageNumber/pageSize are omitted, server-side defaults apply. Use totalCount to decide whether to fetch more pages.
list_send_form_lists
ChatGPTLists batch contact-form requests for the acting subscription. AGENTS: Results are paginated; when pageNumber/pageSize are omitted, server-side defaults apply. Use totalCount to decide whether to fetch more pages.
list_send_form_templates
ChatGPTLists reusable contact-form templates for the acting subscription. AGENTS: Results are paginated; when pageNumber/pageSize are omitted, server-side defaults apply. Use totalCount to decide whether to fetch more pages.
list_talk_scripts
ChatGPTLists talk scripts (phone-call templates) for the acting subscription. AGENTS: Results are paginated; when pageNumber/pageSize are omitted, server-side defaults apply. Use totalCount to decide whether to fetch more pages.
pause_sequence
ChatGPTPause a running sequence. Pending approaches stop, running advertisements are paused and the sequence's email, form and SNS lists are disabled. Resume it with StartSequence. AGENTS: pausing is reversible and loses nothing. Resuming with StartSequence re-enables the sequence's email, form and SNS lists, and approaches queued while it was paused are re-submitted by the batch. Confirm with the user first, because delivery stops while the sequence is paused.
remove_companies_from_company_list
ChatGPTRemoves companies from an existing user company list. AGENTS: companyIds not present in the list are ignored server-side, so it is safe to re-send. The updated list (with new companyCount) is returned.
search_company_v2
ChatGPTSearchCompanyV2 searches companies using structured filters: company attributes, intent, first-party (visitor) intent, domain intent, and saved company lists. All filters are optional; only build the ones the user asked for. AGENTS: WORKFLOW — MANDATORY: Before calling SearchCompanyV2, inspect the requested filters. For every requested field listed below, you MUST call its lookup tool first. - intentFilter.segmentIds / keywordIds: MUST call get_all_targets. Set read_mask.paths to include "id", "name", "keywords.id", "keywords.name", plus every field required to match the user's criteria. Filter the returned targets, use target.id as segmentIds and matching keyword.id values as keywordIds. Never invent or guess IDs. - domainIntentFilter.domainIds: MUST call get_domain_list and pass each matched entry's domainId value (not its id). - visitorIntentFilter.domains / webpageTypes: MUST call get_visitor_domains_and_webpages. Pass domain values to domains and intentPageType values to webpageTypes. Do not pass webpage URLs. - companyListFilter.userCompanyListIds / excludedCompanyListIds: MUST call get_user_company_lists and pass the matched list IDs. - companyFilter.industryIds / excludeIndustryIds: MUST call get_industries and match the Japanese industry name. Use a main category's id (industries[].id) when the user means the whole category; it automatically includes all of its sub-categories. Use a sub-category's id (industries[].children[].id) to match only that sub-category. Do not list a main category's children yourself. - companyFilter.prefectureIds / cityIds: MUST call get_prefecture_list via OpenAPI GET /v2/search/prefectures?includeCities=true. Use prefecture.id for prefectureIds and the matched city.id for cityIds. Show the relevant matching prefecture and municipality options to the user and ask which location to use before searching; do not dump the entire location list. For a municipality such as 目黒区, use its city ID only and do not also send 東京都's prefecture ID, which would broaden the search to all Tokyo. - companyFilter.activityTagIds / businessFieldTagIds / commerceTypeTagIds: MUST call get_company_labels first via OpenAPI GET /v2/search/company-labels and pass the returned label IDs. Activity tags (活動タグ) use type COMPANY_LABEL_TYPE_ACTIVITY_TAG, business content / business field tags (分野タグ) use COMPANY_LABEL_TYPE_BUSINESS_FIELD_TAG, and commerce type tags (業態タグ) use COMPANY_LABEL_TYPE_COMMERCE_TYPE_TAG. Put each ID only in the field matching its returned type; an ID in the wrong field, or an unknown or disabled ID, is rejected with INVALID_ARGUMENT. Do not use get_all_targets for these filters; it only resolves intent segments and keywords. - departmentFilter.minDataFetchDate: Pass the inclusive department acquisition-date lower bound directly in YYYY-MM-DD format. Department names, addresses, and known category IDs are also direct filters. Never invent IDs. If a lookup produces no clear match, ask the user for clarification. Only call SearchCompanyV2 after completing every required lookup. Fields not listed above, such as company names, free-text addresses, ranges, round types, corporation registration numbers, establishment ranges, listing status, listing kinds, revenue ranges, employee ranges, capitalization ranges, representative-name availability, PPC spending, fiscal closing months, representative phone, contact form URL, business outline text, excluded names, excluded addresses, and company domains, can be passed directly using the values documented on each CompanyFilter field. Unknown values are rejected with INVALID_ARGUMENT. Numeric range units and inclusive-bound behavior are documented on the corresponding CompanyFilter fields in this tool's input schema. RESPONSE: each company includes, when available, establishmentDate, employeesNumber, employeeRange, capitalization, listing, listingKind, representativeName, revenue, revenueRange (sales scale), closingMonth, ppcSpending, mainIndustryName (主業界), industries, and its activity / business field / commerce type labels. Monetary response values are in ye…
search_department_v2
ChatGPTSearchDepartmentV2 searches departments using structured filters: department attributes, company attributes, intent, first-party (visitor) intent, domain intent, and saved company lists. All filters are optional; only build the ones the user asked for. AGENTS: WORKFLOW — MANDATORY: Before calling SearchDepartmentV2, resolve every requested lookup-backed filter. - intentFilter.segmentIds / keywordIds: MUST call get_all_targets with read_mask.paths containing "id", "name", "keywords.id", "keywords.name", plus every field needed to match the user's criteria. Use target.id as segmentIds and matching keyword.id values as keywordIds. - domainIntentFilter.domainIds: MUST call get_domain_list and use each matched entry's domainId value (not its id). - visitorIntentFilter.domains / webpageTypes: MUST call get_visitor_domains_and_webpages. Use domain values for domains and webpage intentPageType values for webpageTypes. Do not pass webpage URLs. - companyListFilter.userCompanyListIds / excludedCompanyListIds: MUST call get_user_company_lists and use the matched list IDs. - companyFilter.industryIds / excludeIndustryIds: MUST call get_industries and match the Japanese industry name. Use a main category's id (industries[].id) for the whole category; it automatically includes all of its sub-categories. Use a sub-category's id (industries[].children[].id) to match only that sub-category. - companyFilter.prefectureIds / cityIds: MUST call get_prefecture_list via OpenAPI GET /v2/search/prefectures?includeCities=true. Use prefecture.id for prefectureIds and city.id for cityIds. Show the relevant matching options and ask the user which location to use before searching. For a municipality, send its city ID only; do not also send its prefecture ID. - companyFilter.activityTagIds / businessFieldTagIds / commerceTypeTagIds: MUST call get_company_labels via OpenAPI GET /v2/search/company-labels and use the returned label IDs in the field matching each label's type. Never invent IDs. Ask the user for clarification when no clear match exists. Only call SearchDepartmentV2 after completing every required lookup. Department names, addresses, user-provided category IDs, and minDataFetchDate in YYYY-MM-DD format are direct filters and require no lookup. Other companyFilter values (establishment, listing, revenue, employee, capitalization, representative, PPC, closing-month, phone, form URL, business outline, exclusion, and domain filters) are direct values documented on each CompanyFilter field.
search_job_listings
ChatGPTSearchJobListingV2 searches Sales Marker job listings (求人) and returns one page of job summaries with the hiring company. Filters: title and description keywords, occupation (職種), work location, employment type, posting date range, open/closed status, company IDs, excluded description keywords and company names, recruitment phone presence, and dispatch (haken) prediction. All filters are optional and are ANDed together; values inside one repeated field are ORed (exclusions drop a listing when any value matches). Results are sorted by openedDate (newest first), then by job-listing ID (ascending). Results are summaries and do NOT include the job description. Each result's id is the job-listing ID; companyId and companyName identify the hiring company. This RPC does not create company lists. ACCESS: Only subscriptions with the job-listing feature enabled (subscription.canUseJobListingTable) can call this RPC; others receive PERMISSION_DENIED. AGENTS: WORKFLOW — MANDATORY: Before calling SearchJobListingV2 (MCP search_job_listings, OpenAPI POST /v2/search/job-listing/filters), resolve every requested lookup-backed filter. - prefectureIds / cityIds: MUST call get_prefecture_list via OpenAPI GET /v2/search/prefectures?includeCities=true. Use prefecture.id for prefectureIds and city.id for cityIds. Show the relevant matching options and ask the user which location to use before searching. For a municipality, send its city ID only; do not also send its prefecture ID. - occupations: Every job listing belongs to exactly one of the 18 fixed occupation (職種) values listed on SearchJobListingV2Request.occupations; no lookup call is needed. When the user asks for an occupation, show the matching values from that list (or the full list when the request is vague, e.g. "engineers" matches both "エンジニア(web)" and "エンジニア(電気、電子、機械)") and let the user choose before searching. Send the chosen values exactly as written; never translate, shorten, or invent values. For wording that matches no value, use titleKeywords or descriptionKeywords instead. - companyIds: When the user names a company (e.g. "jobs at トヨタ") instead of giving company IDs, MUST call search_company_v2 via OpenAPI POST /v2/search/company/filters with companyFilter.names set to that name, and use companies[].company.companyId of the matched companies as companyIds. When several companies match, show the candidates (company name and address) and ask the user which to use before searching. Never guess or invent company IDs, and do not put a company name in titleKeywords or descriptionKeywords to find a company's jobs. Never invent IDs or values. Ask the user for clarification when no clear match exists. Title and description keywords, excluded description keywords and company names, employment types, posting dates in YYYY-MM-DD format, openOnly, hasRecruitmentPhone, hasHakenPrediction, and user-provided company IDs are direct filters and require no lookup. SHOWING RESULTS: Present each job's title, companyName, location, employment types, and openedDate, and keep each job's id so the user can ask for more. Tell the user the total (totalCount) and whether more pages exist (hasMore). JOB DETAILS: When the user wants the full description or details of a job (e.g. "show the description", "詳細を見せて", or selecting/clicking a job), call get_job_listing (OpenAPI GET /v2/search/job-listing/{id}) with that job's id from jobListings[].id. Pass the job-listing ID, never the companyId. Do not re-run this search to get details, and do not call get_job_listing for every result up front. COMPANY LISTS: To build a target company list, collect the companyId values from the results (deduplicated) and pass them to the existing company-list tools. PAGINATION: Use size/offset with totalCount and hasMore. Request the next page with offset increased by size while hasMore is true, and keep every filter unchanged between pages. offset + size must not exceed 10000; narrow the filters instead of paging past that limit. ERRORS: On PERMISSION_DENIED, tell the user their plan does not include job-listing search and …
search_meetings
ChatGPTSearch meetings by keyword, title, date range, and assignee. Returns meeting metadata and the linked opportunityID; the transcript is not included in the response. AGENTS: To read a transcript, call get_meeting_details with the meeting id and download the returned transcriptURL. Resolve opportunityID via search_opportunities. You can use the meeting id and the meeting type to get further analysis result via get_note and get_properties.
search_opportunities
ChatGPTSearch opportunities by keyword, stage, assignee, and other filters. AGENTS: Use to resolve a meeting's opportunityID into full opportunity details, or to find deals matching a user query. Always include the searchParams object; pass an empty {} to list without filters.
search_person_v2
ChatGPTSearchPersonV2 searches people using structured filters: person attributes, company attributes, intent, first-party (visitor) intent, domain intent, and saved company lists. All filters are optional; only build the ones the user asked for. AGENTS: WORKFLOW — MANDATORY: Before calling SearchPersonV2, resolve every requested lookup-backed filter. - intentFilter.segmentIds / keywordIds: MUST call get_all_targets with read_mask.paths containing "id", "name", "keywords.id", "keywords.name", plus every field needed to match the user's criteria. Use target.id as segmentIds and matching keyword.id values as keywordIds. - domainIntentFilter.domainIds: MUST call get_domain_list and use each matched entry's domainId value (not its id). - visitorIntentFilter.domains / webpageTypes: MUST call get_visitor_domains_and_webpages. Use domain values for domains and webpage intentPageType values for webpageTypes. Do not pass webpage URLs. - companyListFilter.userCompanyListIds / excludedCompanyListIds: MUST call get_user_company_lists and use the matched list IDs. - companyFilter.industryIds / excludeIndustryIds: MUST call get_industries and match the Japanese industry name. Use a main category's id (industries[].id) for the whole category; it automatically includes all of its sub-categories. Use a sub-category's id (industries[].children[].id) to match only that sub-category. - companyFilter.prefectureIds / cityIds: MUST call get_prefecture_list via OpenAPI GET /v2/search/prefectures?includeCities=true. Use prefecture.id for prefectureIds and city.id for cityIds. Show the relevant matching options and ask the user which location to use before searching. For a municipality, send its city ID only; do not also send its prefecture ID. - companyFilter.activityTagIds / businessFieldTagIds / commerceTypeTagIds: MUST call get_company_labels via OpenAPI GET /v2/search/company-labels and use the returned label IDs in the field matching each label's type. - departmentFilter.minDataFetchDate: Pass the inclusive department acquisition-date lower bound directly in YYYY-MM-DD format. Department names, addresses, and known category IDs are also direct filters. Never invent IDs. Ask the user for clarification when no clear match exists. Only call SearchPersonV2 after completing every required lookup. Other companyFilter values (establishment, listing, revenue, employee, capitalization, representative, PPC, closing-month, phone, form URL, business outline, exclusion, and domain filters) are direct values documented on each CompanyFilter field. Person names, job positions, occupations, emails, social-presence flags, integration flags, and user-provided person IDs are direct filters and require no lookup.
search_project_actions
ChatGPTSearches the project actions of the authenticated user's subscription. AGENTS: All filters are optional and combined with AND. Pass companyId whenever the target company is known: it is the only filter that narrows the search efficiently, and companyName is not a substitute for it. The name/text filters (userName, companyName, projectName, projectTypeName, projectPhaseName, projectPhaseStatus, memo) are contains matches; searchText matches if ANY of those columns contains the text. Results are paginated; when pageNumber/pageSize are omitted, server-side defaults apply. Use totalCount to decide whether to fetch more pages.
send_emails
ChatGPTSends intent emails to a company or person list. This creates a batch (send_email_list) and enqueues it for asynchronous delivery: recipients are resolved from the referenced list (or explicit ids), the batch is persisted, and the individual emails are materialized and sent by a downstream worker. AGENTS: This is a WRITE tool that actually sends email — call it only on an explicit user request to send. Provide exactly one recipient source: userCompanyListId, companyIds, userPersonListId, or privateContactIds. senderEmail, title (subject) and content (body) are required. Omit startDate to send immediately; set it to schedule. The response returns the batch id and the number of recipients queued.
send_forms
ChatGPTSends intent contact-forms to a company list. Recipients are resolved from the referenced company list (or the explicit company ids), and each company's contact form is filled and submitted asynchronously. The sender identity used to fill each form (company name, sender name, address, email, phone, etc.) is taken from the acting user's own profile; callers do not supply it. AGENTS: This is a WRITE tool that actually submits contact forms — call it only on an explicit user request to send. Provide exactly one recipient source: userCompanyListId or companyIds. title (inquiry subject), content (message body) and productId are required. At least one tracking URL (urls) is required, and each must be woven into content via its {checkAccessUrlN} placeholder — e.g. urls[0] as {checkAccessUrl1} — or the request is rejected. Only companies that expose a usable contact form are queued, so the number sent may be smaller than the list. Omit startDate to send immediately; set it to schedule. The response returns how many recipients were queued.
start_sequence
ChatGPTStart a CREATING sequence, or resume a PAUSED one. The engine begins adding members and approaching them. Returns FAILED_PRECONDITION with a StartSequenceBlockedDetail when an attached exclusion list is empty (acknowledgeable) or its CSV import is not ready. Also fails when the node structure is invalid, the configured product no longer exists or an Intent Ad creative is still in screening. AGENTS: On START_SEQUENCE_BLOCK_REASON_EMPTY_EXCLUDE_LISTS, confirm with the user and retry with acknowledgeEmptyExcludeLists set to true. START_SEQUENCE_BLOCK_REASON_EXCLUDE_LIST_IMPORT_NOT_READY cannot be acknowledged; wait for the CSV import to finish.
update_company_list
ChatGPTRenames a user company list and/or replaces its memo. Fields left unset keep their current value. AGENTS: at least one of name or memo must be set. Returns ALREADY_EXISTS when the user already has another list with the new name, and PERMISSION_DENIED when the list belongs to another user and is not editable.
update_sequence
ChatGPTUpdate a sequence's settings. Omitted fields keep their current value. While the sequence is CREATING any field can change. Once it has been started, or is waiting for letter approval, only name can change and any other field is rejected.