account_answer_question
ChatGPTAnswers questions about an account by searching and analyzing account record, meeting transcription, and other related data in the time range if specified. Example questions include: - What specific challenges or pain points is the account experiencing? - What are the desired outcomes or business goals the account aims to achieve?
account_create
ChatGPTCreates a new account in Outreach. The current user is automatically assigned as one of the account's assigned users. The current user is set as the owner if no owner is specified in the input. Returns the created account record including its id only. IMPORTANT: Every field value used to create the account MUST be explicitly provided by the user. Never infer, guess, or fabricate field values. If the user has not supplied a value for a field, do not include that field. ALWAYS call input_fields_fetch with resource="Account" and action="Create" before calling this tool. input_fields_fetch returns the full list of available standard fields, their types, and any required custom fields with validation rules. This ensures you know exactly which fields are available and which are required before creating the account. Example: input: name: "Acme Corp" additional_fields: - field_name: "domain" value: "acme.com" - field_name: "companyType" value: "Enterprise" - field_name: "industry" value: "Technology" - field_name: "custom1" value: "NA"
account_delete
ChatGPTDeletes an account in Outreach by its ID. Returns a confirmation message with the deleted account's ID. IMPORTANT: Always ask the user for explicit confirmation before invoking this tool. Do not proceed with deletion without the user's approval.
account_get_by_id
ChatGPTRetrieve an account using its unique identifier and return the corresponding Account object which includes default fields: id, name, company_type, owner, touched_at, industry, default_connection Use the additional_fields param to include additional account fields in the response. For custom fields (e.g. "Current CRM") which are not already included in the additional_fields parameter description or default fields, first call the filter_fields_fetch tool with entity_type="account" to discover relevant custom fields, then pass that ID (ex. "custom8") in additional_fields. Never guess custom field IDs. Only use when you know the corresponding label. Note: This tool returns only summary information (id, name) for the owner. To get full details about the owner, use the user_get_by_id tool with the owner's id.
account_search
ChatGPTSearch for accounts. Default fields returned: id, name, company_type, owner, touched_at, industry, default_connection. Filters (all optional): name (partial or full), owner_ids, team_id, ids, created date, account plan updated date, last contacted date. - Use last_contacted_before/last_contacted_after to find accounts with or without recent sales activity (sales rep engagement over emails, calls, meetings). - Use additional_field_filters to filter by field values with AND logic. Additional output fields: - Use additional_fields to include extra fields in results. For both additional_fields and additional_field_filters, call filter_fields_fetch tool with entity_type="account" first to discover valid field IDs and labels. Never guess field IDs.
account_search_by_external_id
ChatGPTSearch for accounts using id of the associated object in external systems, such as Saleforce or Dynamics. The search will return a list of Accounts. The Account contains the same information returned by account_get_by_id tool.
account_update
ChatGPTUpdates an existing account in Outreach. Only the fields explicitly included in input are changed; fields omitted from input are left unchanged on the account. Returns the updated account (id, name, owner, assignedUsers). IMPORTANT: Every field value used to update the account MUST be explicitly provided by the user. Never infer, guess, or fabricate field values. If the user has not supplied a value for a field, do not include that field. All fields defined on AccountUpdateInput can be set directly by name. For a custom field (customN), call input_fields_fetch with resource="Account" and action="Update" first to discover its label and validation rules. Example: id: 139 input: additional_fields: - field_name: "industry" value: "Technology"
calendar_events_fetch
ChatGPTRetrieve a list of calendar events (meetings) based on optional filter criteria. A calendar event represents a scheduled meeting or appointment on a user's calendar. Use this tool to fetch scheduled meetings which have not yet occurred or to review past meetings regardless of whether they had transcripts or were processed by Kaia's meeting intelligence features. Use the kaia_meeting_fetch tool to retrieve contents about meetings which have already occurred (transcripts, etc). Use this tool to explore meetings from calendar metadata (e.g. scheduled date, source, cancellation status, etc). Notes: - Multiple filters are combined with AND logic. - Use limit to control how many results are returned. Response: - Each event includes the title, time range, associated prospect/account/opportunity, and attendees. - The 'count' field indicates the total number of matching events, which may exceed the limit. When count > limit, state both the total and the number shown, e.g., "Showing 10 of 247 total events." When count <= limit, state only the count, e.g., "Found 3 calendar events."
call_disposition_fetch
ChatGPTReturns the list of call dispositions (call outcome options, e.g. "Connected", "No Answer", "Voicemail") configured for the organization. Use this tool when: - The user asks "what call dispositions do we have?" or "list all call dispositions" - You need to resolve a call disposition name to its ID - The user wants to know the available outcome categories for calls Response: - Each call disposition includes id, name, order, outcome, defaultForOutboundVoicemail, createdAt, and updatedAt fields. - The response 'count' field indicates the total number of call dispositions, which may exceed the number of records returned when a limit is applied.
call_purpose_fetch
ChatGPTReturns the list of call purposes (the intent/reason categories for a call, e.g. "Discovery", "Demo", "Follow-up") configured for the organization. Use this tool when: - The user asks "what call purposes do we have?" or "list all call purposes" - You need to resolve a call purpose name to its ID Response: - Each call purpose includes id, name, order, createdAt, and updatedAt fields. - The response 'count' field indicates the total number of call purposes, which may exceed the number of records returned when a limit is applied.
call_search
ChatGPTFilter-based lookup that retrieves individual voice call log records. Use this tool when the user wants the actual call records — to list, inspect, or filter them by call log attributes (e.g. "my calls from last week", "calls with no answer for the EMEA team", "outbound calls longer than 5 minutes"). SCOPE — call log records only. Two kinds of question belong to OTHER tools: - CONVERSATION CONTENT → use kaia_meeting_search. This tool carries call log metadata only: no transcript, recording content, topics, or summaries, and no notion of what was actually said. Anything about what was discussed, said, asked, or objected to, about sentiment, or about coaching feedback on how a call was handled goes to Kaia. - STATISTICS AND PERFORMANCE → use the analytics tools, such as analytics_fetch. This tool does NOT aggregate, rank, rate, or trend anything; it returns records, not computed figures. Any question about rates, totals, averages, trends, leaderboards, or patterns across a rep, a team, or a time period belongs to analytics — "what is our answer rate", "average call duration", "how have calls trended", "top callers by rep", "analyze calls for my SDRs over the last 2 months and tell me what patterns you see across the team". Do not fetch a page of call records in order to compute such a figure yourself; the analytics tools answer over the full population, this tool returns at most limit rows. Sales-motion phrases — "cold call", "discovery call", "follow-up call" — describe the kind of conversation, not a call log attribute. There is no filter for them here, so do not approximate one. Only limit/offset pagination is supported. Filterable concepts include (non-exhaustive, conceptual only): - call disposition, call purpose, queue - owner / user, team, sequence, sequence step - related prospect, opportunity, account (scope filters) - tags (collection), prospect tags (collection) - platform (call_source), state, call method (user_call_type) - direction (outbound), recorded - created_at (unique date operator set), duration (numeric, seconds) IMPORTANT: The bullets above describe SEARCHABLE CONCEPTS, not literal attribute names. Do NOT pass any of those labels directly as a filter attribute. The actual attribute identifier and the operators it accepts MUST be obtained from filter_schema_fetch with entity_type="call" before constructing any filter — they cannot be inferred from this description. All filters are passed via filters. There are no dedicated named filter parameters. Required workflow (MUST follow in order): 1. MUST call filter_schema_fetch with entity_type="call" FIRST to discover the exact attribute identifiers and the allowed operators for each attribute. Never guess, abbreviate, or transform attribute names. 2. Build each filter using ONLY an attribute identifier and an operator returned by filter_schema_fetch, then pass them in filters (AND logic). 3. Optionally pass field names in additional_fields to include extra output fields not returned by default. filter_schema_fetch with entity_type="call" also lists these selectable field names. Calling with no filters returns all calls visible to the caller (use limit to cap result size). Returns a Call collection. Default fields always returned (do NOT request these via additional_fields): id, answeredAt, callDisposition, callPurpose, completedAt, createdAt, direction, dispositionDisplay, durationSeconds, endReason, from, note, opportunity, outbound,outcome, platform, prospect, prospectNumber, recordingUrl, sequence, sequenceState, sequenceStep, sequenceStepNumber, state, stateChangedAt, tags, task, to, updatedAt, user. Response shape: - count: the TOTAL number of matching calls (server-side, not limited by limit). - calls: the list of call records (up to limit items). When count > limit, state both the total and the number shown, e.g., "Showing 10 of 50 total calls." When count <= limit, state only the count, e.g., "Found 3 calls."
current_org
ChatGPTReturns the authenticated user's organization (single record): - id: the org's numeric ID. - shortname: short identifier of the org (string). - opportunityAssociationDefault: the org's default opportunity-association strategy — one of "manual", "noop", "recent_closed", "recent_closed_lost", "recent_closed_won", "recent_created", "recent_updated". Use this tool when: - The user asks "what's our default opportunity association?" or other org-configuration questions. - You need to know the server-side default for opportunityAssociation before calling task_create (the org default is applied when the parameter is omitted).
current_user
ChatGPTReturns the user id and email of the caller
emails_search
ChatGPTFetches outbound emails for a specific account or opportunity. IMPORTANT: This tool ONLY returns emails sent BY reps, not emails received from prospects. If the user asks about received/inbound emails, use state='replied' to find emails that got replies, then use email_fetch_thread to retrieve the reply content. If a user query does not reference an account or opportunity: 1. DO NOT call this tool with null values for account_id or opportunity_id 2. Ask the user: "Which account or opportunity would you like to search for emails in?" 3. Wait for the user to provide the information before proceeding 4. A tool call will likely be made to map the account or opportunity name from a user to account_id or opportunity_id Only call this tool after the user has provided one of these identifiers. Additionally, use user_id to filter emails sent or received by a specific user. If a user asks about "my emails" or "emails I've sent/received", use the current user id as user_id.
filter_fields_fetch
ChatGPTFetch fields for a specific entity type (ex. account) which can be used as: - filter attributes on that entity's search tool (ex. account_search, opportunity_search, prospect_search) through the additional_field_filters parameter. - additional output fields on that entity's search or get tools through the additional_fields parameter. Use this tool to retrieve up to 250 additional fields which can be used to retrieve the necessary data to answer user queries that can't be answered with built-in filters and args on existing entity search or get tools. ONLY Supported entity types: - prospect - account - opportunity Response format: - 'id' is the internal id used only for API calls. - 'label' is the human friendly name displayed in the UI that users will be familiar with. - 'validation_type' describes the data type (ex. numerical, inclusion, string, date). - 'validation_definition' describes accepted values for 'inclusion' and 'currency' validation_types. For 'inclusion' and 'currency' fields, the filter value must exactly match one of the entries in validation_definition. Example Response: { "entity_type": "account", "count": 2, "custom_fields": [ { "id": "amount", "label": "amount", "description": "The $ value of the opportunity in annual revenue.", "validation_type": "numerical", "validation_definition": null }, { "id": "custom3", "label": "Account Priority", "validation_type": "inclusion", "validation_definition": [ "Low", "Medium", "High" ] } ] }
filter_schema_fetch
ChatGPTFetch the complete filter schema for a specific entity type, including: 1. Available additional custom fields that can be used as filter attributes and output fields 2. Available additional output fields. 3. The allowed standard filter attributes and their operators, grouped by data type (string, number, date, id). Use this tool to discover: - What custom and standard fields exist for filtering and output selection. - What operators are valid for each standard filter attribute. - What validation types and allowed values apply to custom fields. ONLY Supported entity types: - prospect - account - opportunity - task - sequence_state - schedule - ruleset - call - template - custom_object Response contains three sections: - fields: Lists standard fields can be retrieved. - custom_fields: Lists custom_fields with id, label, validation_type, validation_definition. - filters: Groups standard filter attributes by data type with allowed operators and optional instruction for each group. HOW TO USE FILTERS Each filter object has three fields: - "attribute": field id from this response (e.g. "custom8", "engagedAt"). - "operator": comparison operator (must be valid for the attribute's data type). - "value": list of string values. Value rules: - For "is" operator: multiple values use OR logic (e.g. ["High", "Medium"]). - For "between" operator: exactly two values — [min, max] (e.g. ["1", "100"] or ["2025-01-01", "2025-06-30"]). - For all other value-taking operators: a single value in the list (e.g. ["1000000"]). - For no-value operators (isEmpty, isNotEmpty, and all relative-date operators like today, last7Days, etc.): provide an empty list []. Operator rules: - For "string": use "is" for exact matches, "contains" for partial matches, and "startswith" to match the beginning of a string. Determining allowed operators: - For STANDARD filter attributes: use the operators listed in the filters response for the attribute's type group. - For CUSTOM fields: use the validation_type from the fields response to determine allowed operators: - "string": is, isNot, startsWith, contains, notContains, isEmpty, isNotEmpty - "numerical" or "currency": is, isNot, between, gte, lte, isEmpty, isNotEmpty - "date" or "date_time": is, between, gte, lte, isEmpty, isNotEmpty, plus relative-date operators (today, last7Days, last30Days, etc.) - "boolean": is, isNot, isEmpty, isNotEmpty - "inclusion" or "multi_inclusion": is, isNot, isEmpty, isNotEmpty (values MUST exactly match entries from validation_definition, case-sensitive) If a filter group includes an "instruction" field, follow it for additional guidance on how to construct filter values for that type. HOW TO USE ADDITIONAL FIELDS Use field IDs from the fields or custom_fields response (e.g. "custom8", "engagedAt") as values in the additional_fields parameter on search/get tools. Only pick fields needed to answer the user's query. Rules: - Never guess field IDs. Always use this tool first to discover valid IDs. - Both custom_fields and standard_fields from the response can be used for filtering and additional output. HOW TO USE CUSTOM OBJECTS Custom objects are admin-defined data structures unique to the organization (e.g. Campaigns, Webinars), distinct from standard objects like Prospects and Accounts — and distinct from a standard object's custom fields, which are a different thing fetched with entity_type "account" / "prospect" / etc. With entity_type "custom_object" the response is a custom_objects list plus a custom_objects_count, instead of the three sections above, and a custom_objects_instruction you must follow.
input_fields_fetch
ChatGPTFetch the available input fields for a resource create/update action. Returns a ResourceInputSchema with three sections: - standard_fields: typed optional fields (e.g. companyType, industry, assignedTeams). - ref_types: definitions of non-scalar types referenced by standard_fields (e.g. TeamRelationship, UserRelationship — each with its 'id' field). - custom_fields: organization-specific custom fields with validation metadata. Supported resources: Account Supported actions: Create Response structure: resource (str): The resource type (e.g. 'Account'). action (str): The action type (e.g. 'Create'). standard_fields (list): Standard optional fields, each containing: - name: camelCase GQL field name (e.g. 'companyType', 'assignedTeams') - type: concise type string (e.g. 'string', 'int', 'float', 'TeamRelationship', 'list[UserRelationship]', 'list[string]') - required: whether the field is required for the action - description: description of the field ref_types (list): Non-scalar type definitions referenced by standard_fields. Each contains: - name: type name (e.g. 'TeamRelationship', 'UserRelationship') - description: description of the type - fields: list of fields in this type, each with name, type, required custom_fields (list): Custom fields with validation rules, each containing: - name: internal field identifier (e.g. 'custom1') - label: human-readable label (e.g. 'Region') - required: whether the field is required - description: description of the custom field - validation_type: validation kind ('numerical', 'inclusion', 'string', 'currency', 'date') - validation_definition: accepted values for 'inclusion' type, currency code for 'currency' type, etc. Example response: { "resource": "Account", "action": "Create", "standard_fields": [ {"name": "companyType", "type": "string", "required": false, "description": "Company type of the account"}, {"name": "assignedTeams", "type": "list[TeamRelationship]", "required": false, "description": "Teams assigned to the account"} ], "ref_types": [ {"name": "TeamRelationship", "description": "A relationship reference to a Team entity.", "fields": [{"name": "id", "type": "int", "required": true, "description": "ID of the related entity"}]}, {"name": "UserRelationship", "description": "A relationship reference to a User entity.", "fields": [{"name": "id", "type": "int", "required": true, "description": "ID of the related entity"}]} ], "custom_fields": [ {"name": "custom1", "label": "Region", "required": true, "description": "Region", "validation_type": "inclusion", "validation_definition": ["NA", "EMEA", "APAC"]} ] }
job_role_fetch
ChatGPTReturns the list of job roles configured for the organization. Job roles classify an user's role within their company's buying process (e.g. "SDR", "Sales Manager", "AE", "CSM", "Customer Success Manager"). Use this tool when: - The user asks "what job roles do we have?" or "list all job roles" - You need to resolve a job role name to its ID Response shape: - count: the TOTAL number of matching job roles (server-side, not limited by limit). - job_roles: the list of job role records (up to limit items). When count > limit, state both the total and the number shown, e.g., "Showing 10 of 50 total job roles." When count <= limit, state only the count, e.g., "Found 3 job roles."
kaia_meeting_fetch
ChatGPTFetch call and meeting recording information (summaries and/or transcripts) for specific meetings by instance ID. Use this when you already know the meeting instance IDs. BATCHING RULE: - ALWAYS pass all meeting instance IDs in a single call as a list. - CORRECT: instance_id=["id1","id2","id3"] (one call) - WRONG: three separate calls with instance_id=["id1"], instance_id=["id2"], instance_id=["id3"] Do NOT use for searching by topic or keyword — use kaia_meeting_search instead. RETRIEVAL STRATEGY: - For high-level overviews, action items, or follow-up emails: include_summary=true is usually sufficient. - For exact quotes or verbatim discussion: include_transcript=true. - If summaries do not contain the needed information, immediately call again with include_transcript=true. Do NOT ask the user — retrieve automatically in the same turn.
kaia_meeting_search
ChatGPTSearch across recorded calls and meetings captured by Kaia. Returns matching recordings with transcripts, topics, summaries, and associated account/opportunity info. Use this tool to find recordings by keyword, topic, account, opportunity, participant, host, date range, or title. Kaia captures two kinds of recordings: - Voice calls (phone calls): sourceType = "voice" - Meetings (video/screen-share conferences, e.g. Zoom, Teams): sourceType ≠ "voice" When the user says "calls", filter to voice calls. When they say "meetings", filter to non-voice. When the intent covers both (e.g. "conversations", "recordings"), do not filter by sourceType. FILTER FORMAT: Pass a JSON list of filter objects. Each filter has the shape: {attribute, value, operator}. - attribute: the field to filter on (see supported attributes below) - value: list of string values to match - operator: (optional) defaults to "is". Supported: "is", "isNot" (plus date-specific operators for 'meeting.startTime', see below) All filters are combined with AND logic. SUPPORTED FILTER ATTRIBUTES: - 'search' — full-text keyword search across transcripts, notes, action items, meeting title, participant/account/opportunity names. Syntax: Single word: value: ["budget"] Implicit AND: value: ["budget planning"] Explicit OR: value: ["\"budget\" OR \"planning\""] Exact phrase: value: ["\"budget planning\""] - 'transcript.text' — search only within transcript text (more precise than 'search'). Example: [{attribute: "transcript.text", value: ["budget"], operator: "is"}] - 'topic.type' — detected topic name. General topics: 'budgetpricing', 'competitor', 'demotrial', 'discovery', 'eventmarketing', 'legal', 'logistics', 'nextstep', 'organization', 'product', 'purchase', 'salespitch', 'stakeholder', 'support'. MEDDPICC topics (prefix 'salesMethodology:'): 'salesMethodology:champion', 'salesMethodology:competitor', 'salesMethodology:metrics', 'salesMethodology:economicbuyer', 'salesMethodology:decisionprocess', 'salesMethodology:decisioncriteria', 'salesMethodology:paperprocess', 'salesMethodology:implicatepain'. Auto-enrichment: non-built-in names (e.g. "Salesforce") are automatically mapped to custom topic IDs if a match is found. Built-in types 'competitor' and 'product' automatically include all active custom topics of that type. Example: value: ["competitor", "Salesforce"] finds meetings tagged with any built-in competitor topic, all custom competitor topics, and the custom topic named 'Salesforce'. IMPORTANT: always pair a topic.type filter with include: [{attribute: "topic", includeType: "matching"}] to receive the matching topic segments in the results. Do NOT include transcript. - 'meeting.accountid' — account id - 'meeting.opportunityid' — opportunity id - 'meeting.participant.userid' — participant user id (not just host) - 'meeting.host.userid' — host user id - 'meeting.sourcetype' — value: ["voice"], operator: "is" for calls; value: ["voice"], operator: "isNot" for meetings. - 'meeting.startTime' — date/range filter. Use ONLY these operators (no others are valid): "is" — exact date match: value: ["2026-03-10"], operator: "is" "between" — date range: value: ["2026-03-01", "2026-03-05"], operator: "between" Relative operators (value must be ["0"], operator is one of the following exact strings): "last7Days" — past 7 days (use for "this week" or "last 7 days") "last30Days" — past 30 days "last90Days" — past 90 days "lastMonth" — the previous calendar month "thisMonth" — the current calendar month "today" — today only "untilToday" — from the beginning of time until today "ytd" — year to date - 'meeting.title' — filter by keyword in meeting title - 'meeting.duration' - filter by meeting duration in minutes. Supported operators include: "gte" - greater than or equal to "lte" - less than or equal to "between" - duration between value[0] and value[1] All attributes accept multiple comma-separated values. FILTER EXAMPLES: filter: [{attribute: "search", value: ["Amplify OR budget"]}] filter: [{attribute: "topic.type", value: ["competitor", "Salesforce"], operator: "is"}, {attribute: "meeting.accountid", value: ["123"]}] filter: [{attribute: "topic.type", value: ["salespitch"], operator: "isNot"}] filter: [{attribute: "meeting.sourcetype", value: ["voice"], operator: "is"}] filter: [{attribute: "meeting.sourcetype", value: ["voice"], operator: "isNot"}] INCLUDE FORMAT: Controls which artifact types are returned alongside each matching meeting. Only include artifact types relevant to the query to avoid unnecessary data. attribute values: - 'transcript' — transcript segments - 'topic' — detected topics includeType values: - 'matching' — return only artifacts that match the current filter/search INCLUDE EXAMPLES: Required for keyword search only: [{attribute: "transcript", includeType: "matching"}] Required for topic search only: [{attribute: "topic", includeType: "matching"}] Only when searching both simultaneously: [{attribute: "transcript", includeType: "matching"}, {attribute: "topic", includeType: "matching"}] Do NOT include transcript when topic.type is used as a filter — it returns unrelated segments and inflates the response. PAGINATION: Use 'first' (default 10, max 30) to control page size. Use 'after' with the end_cursor from a previous response to fetch the next page.
kaia_scorecard_fetch
ChatGPTFetch a page of Kaia scorecards configured for the organization. Returns a list of Scorecard objects, each including: id: str name: str | None enabled: bool | None autoScoringEnabled: bool | None metadata: creator/lastEditor/publisher (only id and name are returned for each user), createdAtUtc, modifiedAtUtc, publishedAtUtc questions: list of scorecard questions, each with a type discriminator (ScorecardFreeTextQuestion, ScorecardYesNoQuestion, ScorecardRangeQuestion, ScorecardSingleOptionQuestion, or ScorecardMultiOptionQuestion), plus id, order, question, isDefault, state, metadata (creator/lastEditor/publisher, createdAtUtc, modifiedAtUtc, publishedAtUtc), aiInstructions, and (for option questions) options/intents, and (for range questions) rangeMin/rangeMax Note: This tool returns only summary information (id, name) for creator/lastEditor/publisher. To get full details about a user, use the user_get_by_id tool with the user's id. Note: This endpoint does not report a total count. Use offset/limit to page through results; fewer scorecards than limit in the response indicates the last page.
kaia_scorecard_get_by_id
ChatGPTRetrieve a Kaia scorecard using its unique identifier and return the corresponding Scorecard object including: id: str name: str | None enabled: bool | None autoScoringEnabled: bool | None metadata: creator/lastEditor/publisher (only id and name are returned for each user), createdAtUtc, modifiedAtUtc, publishedAtUtc questions: list of scorecard questions, each with a type discriminator (ScorecardFreeTextQuestion, ScorecardYesNoQuestion, ScorecardRangeQuestion, ScorecardSingleOptionQuestion, or ScorecardMultiOptionQuestion), plus id, order, question, isDefault, state, metadata (creator/lastEditor/publisher, createdAtUtc, modifiedAtUtc, publishedAtUtc), aiInstructions, and (for option questions) options/intents, and (for range questions) rangeMin/rangeMax Note: This tool returns only summary information (id, name) for creator/lastEditor/publisher. To get full details about a user, use the user_get_by_id tool with the user's id.
kaia_scorecard_score_search
ChatGPTSearch Kaia scorecard scores — the filled-in coach cards recorded against a meeting, whether scored by a person or by AI. Use this to answer how a rep was rated on a call, what a reviewer answered question by question, or what AI feedback a call received. Provide at least one of instance_id/instance_ids, scorecard_id, target_user_id, scorer_id or score_id; combined criteria are ANDed. Returns a list of ScorecardScore objects, each including: id: str instanceId: str — the meeting instance the scorecard was applied to targetUser: the user being scored (id and name only) card: the scorecard template (id and name only) responses: one entry per question, each with question: the question, with a type discriminator (ScorecardFreeTextQuestion, ScorecardYesNoQuestion, ScorecardRangeQuestion, ScorecardSingleOptionQuestion, or ScorecardMultiOptionQuestion), plus id, order, question, and rangeMin/rangeMax (range questions) or options (option questions) originalQuestion: the question text as it read when the score was recorded note: free-text note left with the answer response: the answer, with a type discriminator — ScorecardSimpleResponse (answer), ScorecardQuestionYesNoResponse (response), ScorecardRangeResponse (score, floatScore), ScorecardSingleOptionResponse (selected), or ScorecardMultiOptionResponse (selectedOptions) suggestionMoments, aiFeedback: populated only when include_ai_suggestions is true scoredBy: who or what produced the score, with a type discriminator — AIService (id, name) for an AI-generated score, or User (id, name) for a human reviewer visibility: level (PUBLIC, PRIVATE_TO_RECEPIENT, PRIVATE_TO_ME, CUSTOM) and, for CUSTOM, the customAllowList of users createdAtUtc, modifiedAtUtc: str suggestedOverallRating: int | None — calculated overall rating for the scorecard Note: card returns only summary information (id, name). To get the full scorecard template with all its questions, use the kaia_scorecard_get_by_id tool with card.id. Likewise use user_get_by_id for full details of targetUser or a human scorer. Note: This endpoint does not report a total count. Use offset/limit to page through results; fewer scores than limit in the response indicates the last page.
opportunity_add_prospects
ChatGPTAssociates one or more prospects with an opportunity in Outreach. This is the ONLY way to add a prospect to an opportunity. Outreach's API rejects a prospects field on opportunity_create/ opportunity_update outright (it errors as a private relationship that cannot be written, regardless of value) -- never attempt that; always call this tool instead. This runs as an asynchronous bulk-action job, not an immediate write: it returns a Batch describing that job's id and state (e.g. 'queued', 'started', 'finished'), not the opportunity or prospect records themselves. Tell the user the association was submitted/ queued, using the returned state -- do not tell them the prospect has definitely been added unless state indicates the job finished. Example: input: opportunityId: 7 prospectIds: ["42"]
opportunity_answer_question
ChatGPTAnswers questions about an opportunity by searching and analyzing opportunity record, meeting transcription, and other related data in the time range if specified. Example questions include: - What business goals is the buyer aiming to achieve? - What specific challenges is the buyer trying to solve?
opportunity_create
ChatGPTCreates a new opportunity record in Outreach. The current user is automatically assigned as the owner if no owner is specified in the input. Returns an opportunity record with only the id field populated; all other fields are not set. IMPORTANT: Every field value used to create the opportunity MUST be explicitly provided by the user. Never infer, guess, or fabricate field values. If the user has not supplied a value for a field, do not include that field. ALWAYS call input_fields_fetch with resource="Opportunity" and action="Create" before calling this tool. input_fields_fetch returns the full list of available standard fields, their types, and any required custom fields with validation rules. This ensures you know exactly which fields are available and which are required before creating the opportunity. Example: input: name: "Q1 Enterprise Deal" closeDate: "2026-05-17T16:48:12.000Z" additional_fields: - field_name: "amount" value: 50000 - field_name: "opportunityStage" value: {"id": 3} - field_name: "custom1" value: "Inbound"
opportunity_delete
ChatGPTDeletes an opportunity in Outreach by its ID. Returns a confirmation response with the deleted opportunity's ID. IMPORTANT: Always ask the user for explicit confirmation before invoking this tool. Do not proceed with deletion without the user's approval.
opportunity_get_by_id
ChatGPTRetrieve a opportunity and associated sales methodology using its unique identifier and return the corresponding Opportunity object including: id: int name: str | None account: Account | None (only id and name are returned) amount: int | None close_date: str | None created_at: str | None health_score: float | None next_step: str | None last_contacted_at: str | None updated_at: str | None assigned_users: list[User] (only id and name are returned for each user) opportunity_stage: OpportunityStage | None owner: User | None (only id and name are returned) primary_prospect: Prospect | None (only id and name are returned) default_connection: DataConnection | None sales_methodology: SalesMethodology | None (only present when include_sales_methodology=True) fields: list[SalesMethodologyField] SalesMethodologyField contains the definition of a sales methodology field and the opportunity's value for that field, and includes: order: int | None label: str | None name: str | None label_short: str | None type: str | None definition: str | None value: str | None (the opportunity's value for this field) use type and definition if set to determine how to interpret the value field. For example, when type is "RELATIONSHIP", definition is "PROSPECT", and value is "1280273", it means the value is id of the prospect resource. To get more details about the resource, use the corresponding get_by_id tool with the id, or other appropriate tools. In this example, use prospect_get_by_id with id 1280273 to get more details about the prospect. Note: sales_methodology is only available from this tool, not from opportunity_search or other opportunity tools. This tool returns only summary information (id, name) for nested objects such as owner, assigned_users, and primary_prospect. The account field contains the parent account's id and name — use account.id (not the opportunity id) for account-scoped lookups. To get full details about the owner or an assigned user, use the user_get_by_id tool with the user's id. To get full details about the primary prospect, use the prospect_get_by_id tool with the prospect's id.
opportunity_search
ChatGPTSearch for opportunities. Default fields returned: id, account, amount, close_date, created_at, health_score, name, next_step, last_contacted_at, updated_at, assigned_users, opportunity_stage, owner, primary_prospect, default_connection. Filters (all optional): ids, name, account_id, owner_ids, is_open, created date range, close date range, health score range. - Use additional_field_filters to filter by field values with AND logic. Additional output fields: - Use additional_fields to include extra fields in results. For both additional_fields and additional_field_filters, call filter_fields_fetch tool with entity_type="opportunity" first to discover valid field IDs and labels. Never guess field IDs. The Opportunity contains the same information returned by opportunity_get_by_id tool, except sales_methodology which is only available via opportunity_get_by_id. Each opportunity result includes the parent account (account.id, account.name). The response includes: - count: the TOTAL number of matching opportunities (server-side, not limited by the limit parameter). - opportunities: the list of opportunity records (up to limit items). When the user asks about health score without specifying exact numbers, use health_score_low and health_score_high to filter: - any mention of health score without a specific range → health_score_low=0, health_score_high=100 to only return opportunities that have a valid health score When filtering by health score range, sort by health_score ascending to show the most critical opportunities first (sort=["health_score"]). A null/missing health_score means the score is unknown/unavailable — NOT that the score is 0 or low. When presenting results filtered by health score, note any records with missing health score values.
opportunity_search_by_external_id
ChatGPTSearch for opportunities using id of the associated object in external systems, such as Saleforce or Dynamics. The search will return a list of Opportunities. The Opportunity contains the same information returned by opportunity_get_by_id tool, except sales_methodology which is only available via opportunity_get_by_id.
opportunity_stage_fetch
ChatGPTReturns a list of opportunity stages with optional filtering. Opportunity stages represent the different phases in a sales process (e.g., 'Discovery', 'Negotiation', 'Closed Won'). Supported filter attributes: - id - name - isClosed Examples: - filter:[] → returns all opportunity stages - filter:[{"attribute": "isClosed", "operator": "is", "value": "true"}] → only stages that are subtypes of 'closed' - filter:[{"attribute": "id", "operator": "is", "value": ["1", "2"]}] → only stages with id 1 or 2 - filter:[{"attribute": "name", "operator": "is", "value": ["Discovery"]}] → only stages with name that exactly matches 'Discovery' Notes: - Partial string matching is not supported for 'name'. - Multiple filters are combined with AND logic (all must match). Response: - Each stage includes id, name, isActive, isWon, and isClosed fields. - The response 'count' field indicates the total number of matching stages, which may exceed the number of records returned when a limit is applied.
opportunity_update
ChatGPTUpdates an existing opportunity in Outreach. Only the fields explicitly included in input are changed; fields omitted from input are left unchanged on the opportunity. Returns the updated opportunity (id, name, closeDate, owner). IMPORTANT: Every field value used to update the opportunity MUST be explicitly provided by the user. Never infer, guess, or fabricate field values. If the user has not supplied a value for a field, do not include that field. All fields defined on OpportunityUpdateInput can be set directly by name. For a custom field (customN), call input_fields_fetch with resource="Opportunity" and action="Update" first to discover its label and validation rules. Example: id: 7 input: closeDate: "2026-09-30T00:00:00.000Z"
prospect_create
ChatGPTCreates a new prospect in Outreach. The current user is automatically assigned as one of the prospect's assigned users. The current user is set as the owner if no owner is specified in the input. Returns the created prospect record including its id only. IMPORTANT: Every field value used to create the prospect MUST be explicitly provided by the user. NEVER infer, guess, or fabricate field values. Do NOT make up field names. If the user has not supplied a value for a field, do not include that field. At least one of first_name or last_name must be provided. ALWAYS call input_fields_fetch with resource="Prospect" and action="Create" before calling this tool. input_fields_fetch returns the full list of custom fields with validation rules. This ensures you know exactly which fields are available and which are required before creating the prospect. HOW TO BUILD THE INPUT — follow these steps in order: Step 1: Set first_name and/or last_name from the user's request. Step 2: Set owner and assigned_users if specified by the user. Step 3: Collect ALL remaining data the user provided (standard fields like title, company, emails, AND custom fields like custom1, custom2) and place them into additional_fields. Custom fields (custom1 through custom150) are always strings. Step 4: Check input_fields_fetch results for any field with required: true. If ANY required field is missing from custom_fields, STOP and ask the user for the value. Step 5: Verify additional_fields is NOT null and NOT empty [] if you collected any values in Steps 3-4. custom_fields handling: Any field whose name starts with "custom" (e.g. custom1, custom2, …, custom150) MUST be placed inside additional_fields — they are NOT top-level parameters. Each entry has a field_name (e.g. "custom1") and a value (always a string for custom fields). Example — user says "Create prospect Jane Smith, VP of Sales at Acme Corp, set custom1 to NA": input: first_name: "Jane" last_name: "Smith" additional_fields: - field_name: "title" value: "VP of Sales" - field_name: "company" value: "Acme Corp" - field_name: "custom1" value: "NA"
prospect_delete
ChatGPTDeletes a prospect in Outreach by its ID. Returns a confirmation message with the deleted prospect's ID. IMPORTANT: Always ask the user for explicit confirmation before invoking this tool. Do not proceed with deletion without the user's approval.
prospect_get_by_id
ChatGPTRetrieve a prospect using its unique identifier and return the corresponding Prospect object including: id: int name: str | None owner: User | None (only id and name are returned) account: Account | None (only id, name, company_type, touched_at, and industry are returned) assigned_users: list[User] (only id and name are returned for each user) campaign_name: str | None engaged_score: float | None engaged_at: str | None touched_at: str | None stage: Stage | None default_connection: DataConnection | None Note: This tool returns only summary information for nested objects. To get full details about the owner or an assigned user, use the user_get_by_id tool with the user's id. To get full details about the account, use the account_get_by_id tool with the account's id. To get multiple prospects by their IDs, use the prospect_search tool with the ids parameter.
prospect_search
ChatGPTSearch for prospects using: - the supported filters: name (partial or full name), ids (list of prospect IDs), account_ids, opportunity_ids, and owner_ids. - additional_field_filters to filter by field values with AND logic. At least one filter (name, ids, account_ids, opportunity_ids, owner_ids, additional_field_filters) SHOULD be provided; calling with no filters returns all prospects in the system. Returns a list of Prospects. Each prospect contains the same information returned by prospect_get_by_id tool. Default fields returned: id, name, owner, account, assigned_users, campaign_name, engaged_score, engaged_at, stage, default_connection. Additional output fields: - Use additional_fields to include extra fields in results. For both additional_fields and additional_field_filters, call filter_fields_fetch tool with entity_type="prospect" first to discover valid field IDs and labels. Never guess field IDs. Both custom_fields and standard_fields can be used for filtering and additional output. The response includes: - count: the TOTAL number of matching prospects (server-side, not limited by the limit parameter). - prospects: the list of prospect records (up to limit items). When count > limit, state both the total and the number shown, e.g., "Showing 10 of 50 total prospects." When count <= limit, state only the count, e.g., "Found 3 prospects."
prospect_search_by_external_id
ChatGPTSearch for prospects using id of the associated object in external systems, such as Saleforce or Dynamics. The search will return a list of Prospects. The Prospect contains the same information returned by prospect_get_by_id tool.
prospect_update
ChatGPTUpdates an existing prospect in Outreach. Only the fields explicitly included in input are changed; fields omitted from input are left unchanged on the prospect. Returns the updated prospect (id, name, owner, assignedUsers). IMPORTANT: Every field value used to update the prospect MUST be explicitly provided by the user. Never infer, guess, or fabricate field values. If the user has not supplied a value for a field, do not include that field. All fields defined on ProspectUpdateInput can be set directly by name. For a custom field (customN), call input_fields_fetch with resource="Prospect" and action="Update" first to discover its label and validation rules. Example: id: 42 input: additional_fields: - field_name: "title" value: "VP of Sales"
ruleset_search
ChatGPTSearch for rulesets in Outreach by filtering on standard field values. IMPORTANT: The total number of rulesets in an org is usually small. Prefer calling this tool with NO filter to retrieve all available rulesets (raise limit if needed) and do any narrowing yourself from the results. Only pass a filter when it is absolutely necessary — e.g. the user names a specific owner, or the result set is large enough that unfiltered retrieval is impractical. Do not filter by default "just in case." Filterable concepts include (non-exhaustive, conceptual only): - ruleset owner - ruleset name (exact match) - created / last-updated timestamps IMPORTANT: The bullets above describe SEARCHABLE CONCEPTS, not literal attribute names. Do NOT pass any of those labels (e.g. "owner", "ruleset name") directly as a filter attribute. The actual attribute identifier and the operators it accepts MUST be obtained from filter_schema_fetch before constructing any filter — they cannot be inferred from this description. Required workflow when a filter IS actually needed (MUST follow in order): 1. MUST call filter_schema_fetch with entity_type="ruleset" FIRST to discover the exact attribute identifiers and the allowed operators for each attribute. Never guess, abbreviate, or transform attribute names. 2. Build each filter using ONLY an attribute identifier and an operator returned by filter_schema_fetch, then pass them in filter (AND logic). Calling with no filters returns all rulesets in the system (use limit to cap result size). Returns a list of Rulesets. Each ruleset contains: - id, name, createdAt, updatedAt, owner (id, name) The response includes: - count: the TOTAL number of matching rulesets (server-side, not limited by the limit parameter). - rulesets: the list of ruleset records (up to limit items). When count > limit, state both the total and the number shown, e.g., "Showing 10 of 50 total rulesets." When count <= limit, state only the count, e.g., "Found 3 rulesets."
schedule_search
ChatGPTSearch for schedules in Outreach by filtering on standard field values. Use this tool whenever the user asks to look up, list, or filter schedules — including simple lookups. Filterable concepts include (non-exhaustive, conceptual only): - schedule owner - schedule name (exact match) - created / last-updated timestamps IMPORTANT: The bullets above describe SEARCHABLE CONCEPTS, not literal attribute names. Do NOT pass any of those labels (e.g. "owner", "schedule name") directly as a filter attribute. The actual attribute identifier and the operators it accepts MUST be obtained from filter_schema_fetch before constructing any filter — they cannot be inferred from this description. Required workflow when a filter IS actually needed (MUST follow in order): 1. MUST call filter_schema_fetch with entity_type="schedule" FIRST to discover the exact attribute identifiers and the allowed operators for each attribute. Never guess, abbreviate, or transform attribute names. 2. Build each filter using ONLY an attribute identifier and an operator returned by filter_schema_fetch, then pass them in filter (AND logic). Calling with no filters returns all schedules in the system (use limit to cap result size). Returns a list of Schedules. Each schedule contains: - id, name, createdAt, updatedAt, owner (id, name) The response includes: - count: the TOTAL number of matching schedules (server-side, not limited by the limit parameter). - schedules: the list of schedule records (up to limit items). When count > limit, state both the total and the number shown, e.g., "Showing 10 of 50 total schedules." When count <= limit, state only the count, e.g., "Found 3 schedules."
sequence_add_prospects
ChatGPTAdds prospects to a sequence in Outreach by creating a batch operation. Returns a BatchRecord containing the batch ID and status. After receiving the batch response, use the batch_get_by_id tool with the returned batch ID to check its progress and report the final state (total, failures, pending) to the user. Required inputs: - prospect_ids: list of prospect IDs to add to the sequence. - sequence_id: ID of the target sequence. Optional inputs allow controlling mailbox, starting template, opportunity association, batch confirmation, and scheduling. IMPORTANT: Prospect IDs and sequence ID must be explicitly provided by the user. NEVER guess or fabricate IDs.
sequence_create
ChatGPTCreates a new sequence in Outreach. Only the following fields can be set on creation: name, description, salesMotion, owner, sequenceType, shareType, schedule, throttleMaxAddsPerDay, and ruleset. The current user is set as the owner if no owner is specified in the input. For any field the user does not supply a value for: if the field has a documented default value, that default is used; otherwise the field is excluded entirely (no value is fabricated). Returns the created sequence record including its id only. IMPORTANT: Every field value used to create the sequence MUST be explicitly provided by the user. Never infer, guess, or fabricate field values. If the user has not supplied a value for a field, do not include that field, unless the field has a documented default (see parameter descriptions below), in which case that default applies automatically when the field is omitted.
sequence_delete
ChatGPTDeletes a sequence in Outreach by its ID. Returns a confirmation response with the deleted sequence's ID. IMPORTANT: Always ask the user for explicit confirmation before invoking this tool. Do not proceed with deletion without the user's approval.
sequence_disable
ChatGPTDisables a sequence in Outreach by its ID. Use this for "disable", "deactivate", or "turn off" a sequence. Clears the sequence's enabled flag, then cascades that flip to everyone enrolled: each active or failed sequence state is moved to disabled, and its scheduled mailings and tasks are torn down (call tasks are kept). Prospects stay enrolled, but that scheduled work is deleted, not held — re-enabling reschedules from where each prospect stands rather than restoring it. The per-prospect cascade is batched, so on a large sequence it finishes shortly after this call returns. No-op if the sequence is already disabled. Fails if an enable of the same sequence is still in flight. Returns the updated Sequence (id, name, enabled). To turn a sequence back on, use sequence_enable.
sequence_enable
ChatGPTEnables a sequence in Outreach by its ID. Use this for "enable", "activate", or "turn on" a sequence. Sets the sequence's enabled flag, then cascades that flip to everyone enrolled: each disabled sequence state is moved to active, so their steps resume being scheduled. The per-prospect cascade is batched, so on a large sequence it finishes shortly after this call returns. No-op if the sequence is already enabled. Fails if a disable of the same sequence is still in flight. Returns the updated Sequence (id, name, enabled). To turn a sequence off instead, use sequence_disable.
sequence_search
ChatGPTSearch for sequences using additional_filters for field-level filtering with AND logic. Returns a list of Sequences. Each sequence contains: - id, name, owner, engagement_score, sequence_steps The response includes: - count: the TOTAL number of matching sequences (server-side, not limited by the limit parameter). - sequences: the list of sequence records (up to limit items). - page_info: cursor pagination info (end_cursor, has_next_page, etc.) When count > limit, state both the total and the number shown, e.g., "Showing 10 of 50 total sequences." When count <= limit, state only the count, e.g., "Found 3 sequences."
sequence_state_search
ChatGPTSearch for sequence states — records that track an individual prospect's involvement and progress through the steps of a specific outreach sequence. Each sequence state ties one prospect to one sequence and captures their current execution state within that sequence (e.g., 'active', 'bounced', 'paused', 'finished'). A sequence state is NOT the same as a prospect stage. Prospect stages (e.g., 'Attempting to Contact', 'Closed - Unresponsive') describe the prospect's overall disposition in the pipeline. Sequence states describe whether a prospect is currently progressing through a specific sequence's steps. Use this tool when: - The user asks about prospects involved in a specific sequence - The user asks who has been contacted, bounced, replied, or is active in a sequence - The user wants to check a prospect's progress through sequence steps - The user asks about outreach activity tied to a sequence Do NOT use this tool to list the org-level prospect disposition stages — use stage_fetch for that. Supports filtering by account, opportunity, or state. Returns matching sequence states and total count. Each SequenceState includes: id: int prospect: Prospect | None sequence: Sequence | None sequence_step: SequenceStep | None state: str | None user: User | None updated_at: str | None Supports pagination via limit and offset parameters.
sequence_states_destroy
ChatGPTDestroy (delete/remove) sequence states in bulk. Sequence states track a prospect's involvement in a specific outreach sequence. Destroying a sequence state removes the prospect from that sequence entirely. Use this tool when: - The user wants to remove prospects from a sequence - The user wants to bulk-delete sequence states by ID Accepts an input with sequence state IDs to destroy. Returns a Batch object representing the bulk operation. IMPORTANT: Always ask the user for explicit confirmation before invoking this tool. Do not proceed with destroy without the user's approval. After receiving the batch response, use the batch_get_by_id tool with the returned batch ID to check its progress and report the final state (total, failures, pending) to the user.
sequence_step_add
ChatGPTAdds a NEW step to an EXISTING sequence in Outreach. The step is always appended as the last step of the sequence; its position cannot be chosen. Use it when the user asks to add, append, or create a step (an email, call, task, LinkedIn, or SMS touch) in a sequence. NEVER guess, infer, or fabricate a sequenceId, templateId, or any other id — look them up first. Templates can be attached to the new step in the same call via templates, or afterwards with sequence_template_add using the new step's id. Returns the affected sequence as {sequence: {id, name, sequenceSteps}}. The newly added step is the LAST element of sequenceSteps; use its id for follow-up calls such as sequence_template_add.
sequence_template_add
ChatGPTAttaches an EXISTING template to an EXISTING step of a sequence in Outreach. This tool does NOT create the template and does NOT create the sequence step. Both must already exist; this only links them. Use it when the user asks to add, attach, use, or assign a template to a sequence step, or to add another template variant (for A/B testing) to a step that already has one. Takes the id of the sequence, the id of the step within that sequence, and the id of the template to attach. NEVER guess, infer, or fabricate a sequenceId, stepId, or templateId. Templated step types include automatic and manual email, LinkedIn connection requests, LinkedIn messages, and manual SMS. Call and task steps take a note rather than a template, so do not attach a template to those. Returns the affected sequence (id, name) as a confirmation only; the newly created sequence template is not part of the response.
sequence_update
ChatGPTUpdates an existing sequence in Outreach. Only the following fields can be updated: name, description, schedule, scheduleIntervalType, salesMotion, tags, transactional, owner, and shareType. Fields omitted from the input are left unchanged on the sequence. Returns the updated Sequence (id, name, owner, tags).
show_task_list
ChatGPTRenders a read-only task list in the MCP Apps UI. Takes exactly the same input as task_search; see the task_search description for arguments and how to build filters. Use show_task_list only when the user wants to see their tasks displayed. Use task_search instead when you need the tasks as data, e.g. to summarize them, reason about them, or pass them to another tool; show_task_list does not return task data for further use. It runs the search itself, so do NOT call task_search alongside it just to render. Results are sorted by due date unless sort is given.
stage_fetch
ChatGPTReturns the list of prospect lifecycle stages configured for the organization. Stages are org-level configuration labels that classify where a prospect sits in the outreach pipeline. They describe the prospect's overall disposition, NOT their enrollment in a specific sequence. Example stage names: - 'Not Started' - 'Not Started - Added to Sequence' - 'Attempting to Contact' - 'In Progress with Rep' - 'Closed - Unresponsive' - 'Closed - Not Interested - Unsubscribe' - 'Closed - Using Competitive Product' - 'Closed - Left Company' - 'Closed - Bad Timing' Use this tool when: - The user asks "what stages do we have?" or "list all prospect stages" - You need to resolve a stage name to its ID for filtering prospects by stage - The user wants to know the available disposition categories for prospects Do NOT use this tool to find which prospects are enrolled in a sequence or to check a prospect's progress through sequence steps — use sequence_state_search for that. Response: - Each stage includes id and name fields. - The response 'count' field indicates the total number of stages, which may exceed the number of records returned when a limit is applied.
task_create
ChatGPTCreates a new task in Outreach. Returns the created task record including its id only. IMPORTANT: Every field value used to create the task MUST be explicitly provided by the user. NEVER infer, guess, or fabricate field values. Required fields: - taskPriority: Relationship to a TaskPriority entity (use task_priority_fetch to get available priorities). - taskTheme: Relationship to a TaskTheme entity. - subject: Relationship to the Prospect entity this task pertains to. - dueAt: Due date/time in ISO 8601 format in UTC (e.g. '2025-01-15T09:00:00Z'). - opportunityAssociation: Strategy for opportunity association (defaults to 'recent_closed'). The current user is automatically assigned as the owner if no owner is specified in the input. ALWAYS call task_priority_fetch and task_theme_search before calling this tool to get the list of available task priorities, the task themes and their IDs.
task_priority_fetch
ChatGPTReturns the list of task priority lookup values configured for the organization (typically Urgent / High / Normal / Low). Each entry has an id and a name. Use this tool when: - The user asks "what task priorities do we have?" or "list task priority options". - You need to resolve a task priority name (e.g. "Urgent") to its ID for filtering tasks with task_search — the taskPriority filter attribute expects priority IDs, not names. Response: - count: total number of task priorities configured for the organization. The tool returns the first 30 of them, which is every priority for any realistic org; if count is larger than the list, the remainder is not retrievable through this tool. - taskPriorities: list of records, each with id and name.
task_search
ChatGPTSearch for tasks (calls, emails, action items, in-person, LinkedIn, SMS) by filtering on task attributes. Use this tool whenever the user asks to look up, list, or filter their tasks or any other user's tasks (e.g. "my open calls", "overdue tasks for the EMEA team", "tasks scheduled for tomorrow"). Filterable concepts include (non-exhaustive, conceptual only): - task state (incomplete, complete, skipped) and isTrashed flag - owner / assigned user - related account, opportunity, prospect, sequence, stage, sequence step, sequence state - type (action_item, call, email, in_person, linkedin, sms) - origin (manual, sequence, reminder, emailOpen, emailClick, trigger) - task priority, theme, teams - dueAt, completedAt, createdAt, updatedAt, engagedAt, scheduledAt dates - tags (collection), note (text), localTime (numeric) - callDisposition (existence-only) IMPORTANT: The bullets above describe SEARCHABLE CONCEPTS, not literal attribute names. Do NOT pass any of those labels directly as a filter attribute. The actual attribute identifier and the operators it accepts MUST be obtained from filter_schema_fetch with entity_type="task" before constructing any filter — they vary by tenant and cannot be inferred from this description. All filters are passed via filters. There are no dedicated named filter parameters. Required workflow (MUST follow in order): 1. MUST call filter_schema_fetch with entity_type="task" FIRST to discover the exact attribute identifiers and the allowed operators for each attribute. Never guess, abbreviate, or transform attribute names. 2. Build each filter using ONLY an attribute identifier and an operator returned by filter_schema_fetch, then pass them in filters (AND logic). 3. Optionally pass field names in additional_fields to include extra output fields. Only "subject" and "title" are supported; any other name — including custom fields, which are filterable but not selectable here — is an error, so do not pass one. Calling with no filters returns all tasks visible to the caller (use limit to cap result size). Returns a Task collection. Default fields always returned (do NOT request these via additional_fields): id, account, attributableSequenceId, attributableSequenceName, completed, completer, createdAt, dueAt, note, opportunity, owner, prospect, prospectStage, sequence, sequenceStep, state, stateChangedAt, taskCategory, taskType, taskTheme, defaultConnection. Response shape: - count: the TOTAL number of matching tasks (server-side, not limited by limit). - tasks: the list of task records (up to limit items). When count > limit, state both the total and the number shown, e.g., "Showing 10 of 50 total tasks." When count <= limit, state only the count, e.g., "Found 3 tasks."
task_theme_search
ChatGPTReturns task themes configured for the organization — categorization tags that users assign when creating tasks. Each entry has an id, action, and name. Use this tool when: - The user asks "what task themes do we have?" or "list available task themes". - You need to resolve a task theme to its ID for filtering tasks with task_search, or for supplying the taskTheme relationship to task_create. Filtering: Only the following (attribute, operator) pairs are supported in filter; any other attribute or operator is rejected with an error. - attribute: "action", operator: "is", value: ["<action>", ...] - attribute: "id", operator: "is", value: ["<id>", ...] Multiple values within a single filter use OR logic. Multiple filter entries combine with AND. Sorting: sort accepts ONLY these values (prefix with - for descending order): - "id", "-id" - "action", "-action" Response: - count: total number of matching task themes (server-side, not limited by limit). - taskThemes: list of records, each with id, action, and name.
task_update
ChatGPTUpdates an existing task in Outreach. Only the following fields can be updated: taskPriority, taskTheme, subject (the prospect the task is for), owner, dueAt, and note. Fields omitted from input are left unchanged on the task. IMPORTANT: Every field value used to update the task MUST be explicitly provided by the user. Never infer, guess, or fabricate field values. If the user has not supplied a value for a field, do not include that field.
team_get_by_id
ChatGPTRetrieve a team using its unique identifier and return the corresponding Team object including: id: int created_at: str | None name: str | None users: list[User] (only id and name are returned for each user) Note: This tool returns only summary information (id, name) for users. To get full details about a user, use the user_get_by_id tool with the user's id.
team_search
ChatGPTSearch for teams in Outreach. Returns a list of teams matching the given criteria. Supports filtering by favorited by users, content categories, creation date, team IDs, name and teamType for opportunity teams. IMPORTANT: To fetch multiple teams by ID, pass all IDs in the ids parameter as a single call instead of making separate requests for each team. For example, use ids: [1, 2, 3] to retrieve three teams in one request. Returns a list of Teams. The response includes: - count: the TOTAL number of matching teams (server-side, not limited by the limit parameter). - teams: the list of teams records (up to limit items). When count > limit, state both the total and the number shown, e.g., "Showing 10 of 50 total teams." When count <= limit, state only the count, e.g., "Found 3 teams."
template_create
ChatGPTCreates a new template in Outreach. Returns only the id of the created template. Rules for filling in the input: - name (required): MUST come from the user. If the user has not given a name, ask for it before calling this tool. Never make one up, even when you draft the subject and body. - subject, bodyHtml: by default, take them from the user and use their wording verbatim. Draft them yourself ONLY when the user explicitly asks you to write the content, e.g. "create a template named Intro and draft the subject and body for a first email to a new prospect". That permission covers subject and bodyHtml only. When drafting, do not add merge-field variables unless the user asks for specific ones. If the user wants a subject or body but has neither supplied it nor asked you to draft it, ask for it. - bodyHtml should be valid HTML, e.g. "<p>Hi there,</p><p>Thanks for reaching out.</p>". Plain text is also accepted: when the user supplies plain text, pass their wording unchanged. - All other fields (owner, toRecipients, ccRecipients, bccRecipients, shareType, tags, locale, trackLinks, trackOpens, clonedFromId, containsCustomHtml, statsTarget, contentCategoryMemberships) MUST be explicitly provided by the user or resolved from their request via a lookup tool. NEVER infer, guess, or fabricate these values or any IDs; omit a field the user did not ask for. The current user is automatically assigned as the owner if no owner is specified in the input; the call fails if no owner is given and the current user cannot be determined. The template is always created unarchived. Use user_search to resolve an owner's name to a user id, and template_search to resolve the ids of existing templates for statsTarget or clonedFromId. Behavior: - If no owner is given and the current user cannot be determined, the call fails. - The template is always created unarchived.
template_search
ChatGPTSearch for templates in Outreach by filtering on standard field values. Use this tool whenever the user asks to look up, list, filter or compare templates — including simple lookups ("find the Cold Outreach template") and performance questions ("which templates reply best?"). Templates are NOT email-only. The same template entity backs every templated sequence step type: automatic and manual email, LinkedIn connection requests, LinkedIn messages, and manual SMS. (Call and task steps take a note rather than a template, though a call step may carry one that the UI does not surface.) Do not describe results as "email templates" or assume a given template is meant for email. Filterable concepts include (non-exhaustive, conceptual only): - template owner, owning team, the user who favorited it - collection / content category membership - tags - visibility (private, readonly, shared) - archived state - whether the body contains AI-generated personalization content - open / click / reply rate, delivered count - created / last-updated / last-used timestamps IMPORTANT: The bullets above describe SEARCHABLE CONCEPTS, not literal attribute names. Do NOT pass any of those labels (e.g. "owner", "collection", "visibility") directly as a filter attribute. The actual attribute identifier and the operators it accepts MUST be obtained from filter_schema_fetch before constructing any filter — they cannot be inferred from this description. Required workflow when a filter IS actually needed (MUST follow in order): 1. MUST call filter_schema_fetch with entity_type="template" FIRST to discover the exact attribute identifiers and the allowed operators for each attribute. Never guess, abbreviate, or transform attribute names. 2. Build each filter using ONLY an attribute identifier and an operator returned by filter_schema_fetch, then pass them in filter (AND logic). 3. Optionally pass field names in additional_fields to include extra output fields beyond the defaults listed below. Calling with no filters returns all templates visible to the caller (use limit to cap result size). Archived templates: pass the archived filter with value ["true"] when the user asks for archived templates. Do NOT pass ["false"] — omit the filter instead to get the default (non-archived) view. Returns a list of Templates. Default fields always returned (do NOT request these via additional_fields): id, name, subject, shareType, archived, hasAiContent, tags, createdAt, updatedAt, lastUsedAt, owner (id, name), deliverCount, openCount, clickCount, replyCount. Open / click / reply RATES are filterable and sortable but are NEVER returned — they are computed server-side and are not fields on a template. To report a rate, derive it from the returned counts, e.g. reply rate = replyCount / deliverCount * 100. Do not claim a rate when deliverCount is 0. The response includes: - count: the TOTAL number of matching templates (server-side, not limited by the limit parameter). - templates: the list of template records (up to limit items). When count > limit, state both the total and the number shown, e.g., "Showing 10 of 50 total templates." When count <= limit, state only the count, e.g., "Found 3 templates."
user_get_by_id
ChatGPTRetrieve an user using its unique identifier and return the corresponding User object including: id: int name: str | None first_name: str | None last_name: str | None email: str | None teams: list[Team] (only id and name are returned for each team) Note: This tool returns only summary information (id, name) for teams. To get full details about a team, use the team_get_by_id tool with the team's id. To get multiple users by their IDs, use the user_search tool with the ids parameter.
user_search
ChatGPTSearch for users using: - the supported filters: name (partial or full name) and ids (list of user IDs). - additional_filters to filter by field values with AND logic. By default, locked and guest users are excluded (locked=false, guest=false). To include locked or guest users, pass the corresponding filter with value ["true"] in additional_filters. PREFERRED over calling team_get_by_id repeatedly: when you need users from multiple teams, use a single user_search call with a team filter (e.g. {"attribute": "team", "operator": "is", "value": ["1", "2", "3"]}) instead of calling team_get_by_id for each team. Returns a list of Users. Each user contains the same information returned by user_get_by_id tool. The response includes: - count: the TOTAL number of matching users (server-side, not limited by the limit parameter). - users: the list of user records (up to limit items). When count > limit, state both the total and the number shown, e.g., "Showing 10 of 50 total users." When count <= limit, state only the count, e.g., "Found 3 users."
account_get_by_id
Claudeaccount_search_by_ext_id
Claudeaccount_search_by_name
Claudeemails_search
Claudefetch_kaia_meetings
Claudeopportunity_get_by_id
Claudeopportunity_search_by_ext_id
Claudeopportunity_search_by_name
Claudeprospect_get_by_id
Claudeprospect_search_by_ext_id
Claudeprospect_search_by_name
Claudesequence_search_by_name
Claudeuser_get_by_id
Claudeuser_search_by_ext_id
Claudeuser_search_by_nam
Claude