create_call_snippet
ChatGPTCreate a native snippet share and schedule its existing processing task. The URL confirms native creation, not completed media processing or email delivery. share_id is what revoke_share takes. Read list_call_shares for status. Viewer restrictions and expiration behave exactly as whole-call shares; no processing/retry workflow is added.
create_call_tag
ChatGPTCreate an organization-wide tag using the app's tag settings.
create_share
ChatGPTCreate one native call share and optionally send its native email. Returns the share URL and the share_id that revoke_share takes. Without allowed_viewers, anyone holding the link can access it until expiration/revocation. Native notification errors can occur after persistence; do not assume a failed request created no share.
edit_tag_definition
ChatGPTReplace a tag definition. Omitted settings reset to their defaults. Supply the complete intended definition, including any existing description, group, color, and auto-tagging setting that should be retained.
get_agent
ChatGPTRead an agent using the app's live definition permissions. Personal view accepts the native workflow/template/installation identifiers. Organization view requires a template_id from the organization directory and returns that same safe projection. Neither view installs or executes agents.
get_agent_run
ChatGPTRead the full latest output of one of your Airspeed agent runs. Accepts any run list_agent_runs can return: a chat or a run of a saved agent. For a conversation, the output is its latest reply. This retrieves an existing result; it does not start or rerun an agent. Use url_link to cite the run or send the user to it. Only runs owned by the signed-in user in their organization are available, including for admins. Missing runs and other people's runs both answer "Agent run not found".
get_agent_thread
ChatGPTRead native conversation detail by exactly one thread ID or share ID. Thread IDs require ownership. Shared reads recheck the live share, tenant, and underlying resource access through the same app dependencies. Returns the app's visible messages and metadata, including native pending approvals for owner reads; this does not make approval decisions or start an agent.
get_call_activity_stats
ChatGPTRead native call-count, duration and talk statistics. Users with activity-page permission default to the organization cohort; others default to their own home-widget statistics and cannot select others. Private calls retain the app's numeric aggregate treatment, without returning call details. Dates use the native unbounded start/current-time end; exact time-range filters are not offered because the native API cannot honor them. Every request scans the full call history in scope, so avoid repeated calls.
get_call_info
ChatGPTRead a call's summary, extracted insights, and transcript. Also returns participants, companies, tags, an Airspeed call link, and recording details when available. Stored CRM deal, contact, lead, and company IDs are included when available, separately from Airspeed IDs; each company's id is its Airspeed company ID, and a participant's user_id is their Airspeed user ID when matched. Use this to answer questions about what was discussed, building on Airspeed's existing analysis. Fields may be absent when their data is unavailable. Use url_link to cite the call. Only calls the user may read in full in Airspeed are available, including private calls they participated in. Missing, redacted, and hidden calls all answer "Call not found"; this does not prove a call never happened. Recording URLs expire after 24 hours. Set transcript={} to request bounded native raw text, or supply language for native translated turns. Truncated means content past the bound is not available over MCP: there is no offset or cursor, so repeating the request returns the same prefix. Raise the bound (up to 20000 characters or 100 turns) or send the user to url_link for the rest. A translated request translates the complete transcript every time, so do not repeat it. The default inline inferred transcript and recording behavior remains unchanged. comments_limit optionally returns recent top-level comments in native order, with direct reply counts but no reply bodies. reply_count does not include replies nested under those replies. The native reader still loads the whole comment tree; truncated reports omitted top-level comments, not a cursor.
get_call_media
ChatGPTGet a call's audio or video recording link and media type. Use this when the user wants the recording. For a transcript, summary, or insights, use get_call_info instead. Recording URLs expire after 24 hours; media may be absent when no recording is available. Only calls the user may read in full in Airspeed are available, including private calls they participated in. Missing, redacted, and hidden calls all answer "Call not found".
get_call_snippets
ChatGPTRead existing clips from a call, with their transcript excerpts. Returns saved snippets, newest first, with start and end times and recording links when available. Transcript turns cover only each snippet's range. This does not create clips or search for new excerpts. Use get_call_info for the full transcript. Recording URLs expire after 24 hours. Only calls the user may read in full in Airspeed are available, including private calls they participated in. Missing, redacted, and hidden calls all answer "Call not found". No snippets does not mean no call transcript.
get_changelog
ChatGPTRead what changed in these tools, as Markdown, newest change first. Covers removed, renamed and merged tools and changes to their inputs or results that a client may need to act on. Call it when a tool you expected is missing or a call that used to work is rejected. The live tool list remains the source of truth for current schemas.
get_company
ChatGPTRead a company with optional native app sections, loaded only on request. Includes: summary, people, people_summaries, hierarchy (ancestors/direct children), deals, similar_companies, crm_fields, activity, slack_threads. Omitted sections are null. crm_fields reads the organization's live CRM. Use get_company_profile for the CRM and Crunchbase profile and get_company_emails for email threads. List sections preview at most 100 top-level entries. truncated means only that MCP trimmed the native response; nested content and upstream reads retain native limits. No section continuation or date filters exist. Activity/messages may be incomplete because native sources limit results or omit failures. Activity's default 90-day window applies only to CRM activity. Null CRM fields retain the app's unavailable semantics.
get_company_emails
ChatGPTRead email threads with a company from the CRM and connected mailboxes. Returns up to 100 threads from the organization's CRM and from the Google and Microsoft mailboxes of org users who allow email reading. Calls those providers live; an expired mailbox token is refreshed and stored, and a token that cannot be refreshed is removed, so it is not read-only. Results may be incomplete because providers limit results or omit failures.
get_company_profile
ChatGPTRead a company's CRM and Crunchbase profile for a CRM-linked company. Returns firmographics from the CRM account and from Crunchbase (name, description, industries, headcount, revenue, funding, address, links), recent news, and funding rounds. Requires an existing CRM link on the company; no association discovery runs. Calls the CRM and Crunchbase live and may refresh Airspeed's stored provider caches, so it is not read-only. A null profile means no source returned data, which is not distinguishable from a provider failure.
get_deal
ChatGPTRead a visible CRM deal and optional native stored sections. Includes summary, insights, edit_options, field_history, and close_probability_history. Omitted sections are null. Native feature gates apply. edit_options requires edit permission and contains only choices for allowed edits. Field history is newest first; probability history is oldest first within its latest readings. history_limit applies to each requested history independently, with no continuation or full CRM audit guarantee. A closed-lost deal's loss_reasons hold Airspeed's inferred reasons, each with its reasoning and matched loss-reason category.
get_forecast_periods
ChatGPTFind labeled forecast periods using your organization's fiscal calendar. Returns periods from two years ago through two months ahead. Use their start and end dates when reading or submitting forecasts. Forecasting and forecast boards must be enabled and the caller must have access to forecasting.
get_forecast_rollup
ChatGPTRead a forecast board's rollup for your direct reports, as in Airspeed. Returns one row per direct report with their submitted call and per-category amounts, the amounts derived from their board deals, deal counts, quota, and pipeline coverage, plus the reconciliation totals, category order, and currency. Rows are empty when the viewer has no direct reports; use list_forecast_history for an individual's submissions. Use get_forecast_rollup_deals for the deals behind a row.
get_forecast_rollup_deals
ChatGPTRead the open board deals behind a forecast rollup, grouped by owner. Returns the deals Airspeed shows in the rollup's review drawer: each group's open deals with whether the rep selected them, open deal and closed-won counts, plus the viewer's saved inclusion changes and deal adjustments. Only deals you can currently read in your CRM are returned.
get_forecasting_settings
ChatGPTRead the fiscal calendar, period type, and forecast category overrides. Use editable_on_submit to determine which categories accept submitted values. This returns forecast configuration only. Forecasting and forecast boards must be enabled and the caller must have access to forecasting.
get_loss_reasons
ChatGPTCount closed-lost deals by Airspeed's inferred loss-reason category. The strategic page's loss-reasons chart. Airspeed infers each lost deal's reasons and matches them to the organization's loss-reason categories and subcategories, most frequent first. Requires the strategic page's view-others-calls permission. close_date defaults to the last 90 days and is compared as the app does: the start date inclusive, the end date exclusive. Deals that close after midnight on the last included day are left out, as in the app. Counts cover every lost deal in the cohort; crm_deal_ids list only the deals you can see, for get_deal, which returns each deal's reasoning. Each inferred reason counts once, so one deal can count toward several categories; total_lost_deals counts deals. Amount bounds must be positive integers. Demo reports use existing fixtures, report filters_applied=false and return a null close_date because no window was applied.
get_meeting_recording_status
ChatGPTRead a past meeting's recording outcome using IDs from your calendar. The bot must be a persisted CallBot belonging to your organization, so a bot another organization sent to a joint meeting is not readable. The exact pair must still appear on your current native calendar list. This reloads the native list before requesting the ended meeting's status; a stale, missing, or mismatched pair is rejected. Native calendar read limits and costs apply.
get_playbook
ChatGPTRead the latest content of a playbook in your organization. Use this when the user asks about a playbook's guidance. To read an older version, find it with list_playbook_versions and use get_playbook_version.
get_playbook_version
ChatGPTRead a specific version of a playbook in your organization. Use this for historical guidance or to compare versions. For the latest content without choosing a version, use get_playbook.
get_scorecard
ChatGPTRead a current organization scorecard definition. Its rubric is current, not a historical snapshot of the scale used for previously scored calls.
get_scorecard_report
ChatGPTRead the native team overview with its existing overview entitlement. Optional users selects self or Airspeed IDs; view selects everyone/my_team inside current ScorecardScope. This report uses native numeric aggregates, not full call detail. Rubric scales are current, not historical snapshots. No revenue lookup, scoring, cleanup, or report persistence is performed. Date filters are omitted because the native report's raw timestamp query does not implement the shared date-range semantics. Use per-user history for supported calendar-date filtering. No cursor or exhaustive call claim.
get_strategic_insights
ChatGPTRead the app's native strategic insight chart and deal-outcome report. Requires the strategic page's view-others-calls permission. Native aggregate cohorts and access-filtered deal references stay intact. The app defaults to its recent period and expands chart dates to week/month buckets; deal outcomes have a lower date bound, not the chart's upper bound. No exact period filter is offered; get_loss_reasons reports loss reasons. Amount bounds must be positive integers. Demo reports use existing fixtures and explicitly report filters_applied=false.
invite_user
ChatGPTInvite one person to your organization using the app's invitation flow. Invites a single person per call; call it once for each additional person. Requires organization-settings permission. Native seat availability, forecasting-seat and owner-role checks apply. This creates an invitation, attempts its email notification, and records its email domain. A successful response is the app's acceptance, not an email-delivery confirmation. Failures may occur after persistence: inspect organization users before retrying. Support and flag-gated restricted roles are not offered here.
join_meeting
ChatGPTSend the recording bot to a meeting URL now, subject to app permissions. Native recording-seat, organization call-cap and recent-bot deduplication apply. created confirms a bot was created and stored, not that it joined successfully. No future recording preferences are changed.
list_agent_runs
ChatGPTList your Airspeed agent runs: chats and runs of saved agents. Mirrors the app's Recent activity. With no filters you get its All tab: every agent conversation you started and every run of your saved agents, whether it ran manually, on a schedule or on an event, newest by start time first. Each run's kind is chat (no saved agent) or agent, and an agent run's trigger_type says how its agent starts. To find the agents themselves rather than their runs, use list_agents. Returns titles, output previews and a url_link that opens each run in the app; use it to cite a run. Use get_agent_run with a returned run ID to read the full latest output before summarizing it, get_agent_thread for the whole conversation, or run_agent with the run ID as thread_id to continue it. This tool reads existing runs; it does not start agents. Only runs owned by the signed-in user in their organization are returned, including for admins. No email or user ID filter is needed for "my runs". Prefer the typed filters object for new clients. Legacy flat selectors remain supported, but cannot be mixed with filters. Each ID-set selector accepts exactly one ID because the native API has single selectors. Arbitrary sorts are absent.
list_agent_threads
ChatGPTList your conversations using the app's complete, unpaged native list. Only your own threads are returned, including for administrators. The only filter is origin; order and content visibility follow the app. Use list_agent_runs for a paged, filterable run history, or get_agent_thread for full detail.
list_agents
ChatGPTList the app's native personal agents or safe organization directory. These are complete, unpaged native lists with their native ordering. Personal view includes owned/created workflows and supports subscriber_agent_ids.any_of with exactly one ID. Organization view uses the app's safe discovery projection, not other people's installation settings. Use returned IDs with get_agent in the same view. Native personal configuration may resolve Slack notification metadata. agent_type splits either view like the app's Manage page: "live" agents have a schedule or event trigger, "on_demand" agents run only when asked.
list_call_shares
ChatGPTList the call's native active shares, newest first and unpaged. Requires full call visibility. Expired history, viewer audit logs and raw storage URLs are not returned. Snippets include current processing status.
list_call_tags
ChatGPTFind the call tags available in your organization. Returns tag IDs, names, descriptions, and groups, oldest first. Use this to translate a user's tag name into tag_ids for list_calls; do not invent IDs. Listing a tag does not grant access to calls carrying it.
list_calls
ChatGPTFind calls by participant, date, title, or tag, newest first. Returns previews with call IDs, titles, participants, companies, dates, and tags. Stored CRM deal, contact, lead, and company IDs are included when available; these are CRM record IDs, not Airspeed call IDs. Each company's id is its Airspeed company ID, and a participant's user_id is their Airspeed user ID when matched. Use get_call_info for summaries, insights, and transcripts. For "my calls", filter participant_email to the user's confirmed meeting email. Use list_users to look up names and the authenticated account. Ask if the meeting email is still unclear; do not guess it or assume the client's account email is their meeting email. Without this filter, results include all calls the user can access, not just calls they attended. Only calls the user may read in full in Airspeed are returned, including private calls they participated in. Redacted and hidden calls are omitted. An empty result means no accessible calls matched these filters. Optional filters groups title, participant_email and tag_ids.any_of using the same existing native selectors. Do not mix it with title_filter, participant_email or tag_ids. It has no dates; combine it with start_time_from and start_time_to for a date range. Exact timestamp filters/sorting beyond the native legacy behavior are not supported; omitted and unsupported selectors must not be guessed.
list_companies
ChatGPTFind accessible companies by name/domain text and native app ordering. Text matches name, primary domain, or additional domains. Sort by domain, last_call_at, or next_call_at; default last_call_at descending. Native offset pagination can shift as records change. Keep filters, sort, and limit unchanged when passing the returned cursor. Only these native filters are supported; use get_company for optional detail sections.
list_deals
ChatGPTList accessible CRM deals with native stage, owner, and closed filters. stage_ids.any_of accepts one ID from get_deal's edit_options; owner_ids.any_of contains CRM owner IDs from returned deals, and owner_user_ids.any_of takes Airspeed user IDs instead; together they match any listed owner. All supplied filters combine. Native order is retained. Keep filters unchanged with the returned cursor. limit is a native page target: Salesforce can return more readable rows; none are discarded. Continue while next_cursor exists, including after empty pages. Date/text filters are unsupported. Native enrichment can omit failures.
list_forecast_boards
ChatGPTFind forecast boards and the deal filters they apply. Use a returned board ID for board-scoped reads and submissions. Both forecasting and forecast boards must be enabled. Forecasting access is required, except that organization settings administrators may read configuration without a forecasting seat, as in Airspeed.
list_forecast_history
ChatGPTRead forecast submissions and their saved deal snapshots, newest first. Defaults to your own history. Managers may request users in their reporting chain; users with full forecast visibility may request any user in their organization. Requests containing inaccessible users fail. Returns the same history values as Airspeed, including saved deal snapshots, calculated manager totals, and fallback quotas. Without board_id, returns every submission, including ones made before forecast boards existed; those have board_id null and belong to no board. With board_id, returns only that board's submissions and leaves out pre-board ones. Entries made on a board include its board_name unless the board has since been deleted.
list_meetings
ChatGPTRead your native calendar or one permitted direct report's calendar. Select users.mode=self (default), or users.mode=ids with one Airspeed ID. Direct-report responses omit private meetings, join links and attendee emails, exactly as in the app. The native response is unpaged; no additional date, text or recording filters are offered. Recording choices, bot IDs, linked call IDs and attendees with their Airspeed user IDs are available on your own calendar rows.
list_playbook_versions
ChatGPTFind previous versions of a playbook in your organization. Returns version previews, newest first. Use get_playbook_version to read a selected version's content or compare versions. For the latest content alone, use get_playbook.
list_playbooks
ChatGPTFind your organization's playbooks before reading their content. Returns paginated previews, newest by playbook creation time first. Use get_playbook with a returned ID to read the latest version's content. Deleted playbooks are excluded. updated_at reflects the latest version's update time, falling back to its creation time or the playbook's creation time if no version exists; results are not sorted by updated_at.
list_scorecard_results
ChatGPTRead native per-call results, one user's history, or one scorecard's range. Choose filters.call_ids={any_of:[one Airspeed ID]}, or filters.users={mode:self|ids,ids:[one Airspeed ID]}, or filters.scorecard_ids={any_of:[one Airspeed ID]} with started_at. User history permits scored_on calendar-date gte/lt bounds mapped to native inclusive UTC day buckets; it returns at most 100 rows per scorecard and has no cursor. Per-call mode returns a result list; history keeps the native user/grouped-card response. Scorecard range mode pages every readable user's results for that scorecard whose call started in started_at (ISO datetimes with offsets, gte inclusive, lt exclusive, both required, at most 92 days apart), newest first, with limit and an opaque cursor from pagination.next_cursor. Rows are history rows plus user_id and user; a call whose detail is hidden keeps its scores, drops feedback and call_context, and is placed at the start of its UTC day. Rows omit skill_example. Current ScorecardScope and CallScope apply. Native history may retain numeric scores when call detail is hidden. max_score and skill metrics use the current rubric, not a historical scale; removed rubric skills may be absent. Do not infer historical comparability, exhaustive coverage, or missing scores as zero.
list_scorecards
ChatGPTList native scorecard definitions in the current organization. Uses the app's definition list, including its current rubric, assignments and aggregate average. enabled_only selects the app's enabled-list operation. This native list has no cursor or additional filters.
list_tasks
ChatGPTRead your assigned inbox, or tasks for one visible call or company. Use filters.call_ids.any_of or filters.company_ids.any_of with exactly one ID to select the native source. Company tasks require Companies access. Source-call visibility is checked before returning task content. Standalone inbox tasks remain available. Native sources load complete lists in native order: limit only bounds the response, and truncated reports omitted visible tasks. There is no cursor, status/date filter, or source query bound. Company labels are omitted.
list_users
ChatGPTList organization users with their manager, seat and invitation details. Returns one complete, unpaged list, always sorted by name then email. Omitting filters or passing {} lists active users without team narrowing. include_invites adds pending invitations; managed_by selects the manager and their recursive reports. Invitations retain organization-wide scope even when managed_by is supplied; invitations of active users outside the team are still omitted. Use names and emails to find people for participant_email in list_calls. Accepted users carry their Airspeed user id; pass it to user tools, never an email or CRM owner ID. Rows with has_accepted_invite false carry an invitation id, not a user id; do not pass it to user tools. Invitation names are empty strings. is_current_user identifies your authenticated Airspeed account. role is the Airspeed access role; job_title is the person's position, if set. Follow manager_id or manager_email through the user list to understand the reporting chain; manager fields are null if no active, non-support manager is in the user list. team_label is the user's configured team label, if set. Excludes departed users and support accounts. Listing a person does not grant access to their calls or other records.
remove_meeting_bot
ChatGPTRemove a caller-created ad-hoc bot using the app's recording permission. Only a persisted CallBot with your user and organization is supported. Calendar participation or a colleague's shared bot does not authorize removal. Use the recall_bot_id returned when you created it with join_meeting; a reused colleague bot remains unsupported. Native leave polling waits for confirmation; failures can have unknown outcome.
revoke_share
ChatGPTRevoke a native call share after checking full access to its parent. In-progress snippets cannot be revoked. Native storage deletion can precede the database change, so unexpected errors have an unknown outcome.
run_agent
ChatGPTStart or continue a one-off Airspeed agent conversation. This runs an interactive agent, not a scheduled or event-triggered job. Use list_agent_runs / get_agent_run to read those outputs. Omit thread_id to start a new conversation. Always pass the returned thread_id on later calls so follow-up questions keep the same context. Starting again without it creates a new conversation.
search_airspeed
ChatGPTSearch calls, companies and deals, keeping each source's native order. Results are grouped in requested resource order with at most limit per source. This is a bounded native window, not exhaustive search or cursor pagination. Current access gates may further reduce it. Call filters require resources=['calls']; no equivalent company/deal filters are exposed. Unsupported timestamps, upper dates, has_deal, and arbitrary duration ranges are rejected rather than approximated. Native call search uses Atlas; company and deal search use their existing app paths. A source that fails is left out and listed in unavailable, so an empty data list is not proof of no matches; other requested sources still answer, and the tool fails only when every requested source fails. Help-center articles are searched by search_support_docs.
search_support_docs
ChatGPTKeyword-search published Airspeed help-center articles. Returns up to limit (at most five) articles in the help center's native relevance order, each with a title, URL and short snippet. Articles come from the public help center, not from the caller's organization data, so the result is the same for every caller. A help-center outage is an error, not an empty result.
set_call_tags
ChatGPTAdd or remove one tag on a call you can read in full. This changes a single tag; it does not replace the call's other tags. Adding a grouped tag replaces the existing tag in that group. The app's scorecard follow-ups apply: results whose tag filter no longer matches are deleted, and scorecards filtered on an added tag are scored in the background. Scheduled scoring is best-effort and may finish after this returns. Adding a tag the call already has changes nothing.
share_agent_thread
ChatGPTShare your conversation with your organization, as the app's Share button. Anyone in your organization can open the returned url, or read the conversation with get_agent_thread and the returned share_id. A conversation has one link: sharing it again, including after it was unshared, returns the same link. Ownership is required even for administrators.
stop_agent
ChatGPTRequest that your conversation's in-flight agent work stop. Native accepted=true means the stop request was recorded, not that work has finished stopping. Ownership is required even for administrators.
submit_forecast
ChatGPTSubmit or resubmit your forecast on a board, as in Airspeed. Each call appends a new submission owned by you; earlier ones stay in list_forecast_history, which also shows your existing amounts and comment. This tool cannot submit on another user's behalf. The period must be one of get_forecast_periods' options, copied exactly. Amounts may name forecast (your call) and the categories that get_forecasting_settings marks editable_on_submit; other categories are rejected. closed_won is derived from deals and cannot be stated. A personal forecast lists your deals. A manager forecast also passes forecast_ids: the submission_id of each row in get_forecast_rollup for the same board and period, which are already limited to your direct reports. Rolling up a forecast from another board or period fails. Returned deals carry company_id, the Airspeed company ID, when resolvable.
update_call
ChatGPTApply one native title or ordinary CRM-deal reassignment. Requires full call access; native processing and CRM policies still apply. Pending deal-match choices stay in the app because their replay starts rubric scoring. Native writes have no new concurrency guarantee. Follow-ups are best-effort background work; scheduled does not mean complete. On an unknown outcome inspect current state before retrying.
update_deal
ChatGPTApply supported fields through the native authorized CRM writer. Change amount, close_date (ISO datetime with offset), stage_id, forecast_category, or configured extra_crm_fields. Omitted fields stay unchanged; null does not clear. Native validation rejects amount=0 alone. Stages resolve from current metadata; extra fields retain native writable field checks. CRM record read/edit checks run as the caller. Success confirms the native write returned; cache/history remain native best-effort effects. Follow-ups are scheduled, not confirmed complete. Use get_deal for readback; do not automatically retry an error whose outcome is unknown.
update_forecast_quotas
ChatGPTUpdate forecast quotas with the web app's organization-settings permission. Supply a user_id, nonnegative forecast_quota, and optional forecast period for each update. A period also saves a quota for that specific period; every update sets the user's current quota. Null quota values are skipped. Returns only updated users in your organization; unknown IDs and users outside your organization are skipped. Requires permission to edit organization settings, which administrators hold by default.
get_agent_run
Claudeget_call_info
Claudeget_call_media
Claudeget_call_snippets
Claudeget_playbook
Claudeget_playbook_version
Claudelist_agent_runs
Claudelist_call_tags
Claudelist_calls
Claudelist_playbook_versions
Claudelist_playbooks
Claudelist_users
Clauderun_agent
Claude
