Tools reference

Website domain, e.g. "acme.com". Omit when the account has only one.

One section only.

Metric or dimension names to define in full.

Website domain, e.g. "acme.com". Omit when the account has only one.

Website domain, e.g. "acme.com". Omit when the account has only one.

Metric names. Defaults depend on the first dimension (traffic, outcomes or ad numbers).

Dimensions to break down by, crossed, e.g. ["ad_campaign"] or ["channel","landing_page"].

Add a time column: one row per bucket (per dimension value), in the workspace time zone.

Period in plain language: "last 30 days", "yesterday", "July 2026", or "2026-07-01..2026-07-31". Defaults to the last 30 days.

Start day (YYYY-MM-DD), instead of period.

Last day, INCLUSIVE (YYYY-MM-DD).

Dimension filter, "dimension:operator:value". Operators: eq, ne, in, nin, contains, not_contains. Examples: "channel:in:paid_search,paid_social", "country:eq:US", "campaign:contains:brand". Several are ANDed unless match="any".

Combine filters with AND (all, default) or OR (any).

Attribution models crediting conversions and revenue to the dimensions. Several give one column per model (conversions__linear). Ad dimensions default to last_non_direct.

What counts as an outcome. Defaults to `conversions`. ONE request counts ONE kind: 2 MQLs is not a slice of 39 conversions, it is a different object, so never blend them or present a total across them.

• conversions (All conversions, from tracking): Every conversion your tracking script recorded: form submissions, meetings booked, chats, payments. One total, no split.
• conversion_type (Conversions by type, from tracking): The same conversions, split by what kind they were: form fills, signups, meetings, chats, payments and the rest. Which kinds appear depends on what this workspace records. These add up to the conversions total.
• contact_stage (Contacts by stage, from crm): People who reached each stage of your funnel, from your CRM. Counts PEOPLE, not conversions, so these do not add up to the conversions total. Someone now at Customer still counts once under every stage they passed through.
• deal_status (Deals by status, from crm): Deals from your CRM, with the money on them. Counts DEALS, not conversions. Pipeline is the amount on deals created; Won revenue is the amount on deals closed won.
• subscription_status (Subscriptions by status, from payments): Trials, upgrades to paid, and cancellations from your payment provider, with the recurring revenue each one added or removed. Counts subscriptions, not conversions.
• payment_revenue (Payments, from payments): Individual payments you actually received, through Stripe, Shopify or whichever provider you connected. Subscription renewals are not counted here.
• revenue (Revenue, from payments): Every money movement we hold for this workspace: payments and refunds from your payment provider, won deals from your CRM, values you typed on a contact, and anything your tracking script reported. Deduplicated so the same money is never counted twice, and credited to channels by the selected attribution model.

Call get_context first to learn which of these this workspace has data for, and the stage keys to pass as outcome_values.

Narrow the outcome to specific stages, e.g. ["mql"] or ["deal_won"]. Keys come from get_context. Omit to count every value of that kind, which is what "show me our funnel" means.

WHO the outcome is counted over. Only contact_stage offers companies. The company view is SMALLER than the contact view rather than a slice of it: a company sits at the stage of its most advanced contact, so three people at one account are one company. Use it when the question is about accounts rather than people.

Only count these conversion EVENT names, e.g. ["demo_booked"]. Takes precedence over types.

Only count these conversion categories, e.g. ["Web Form","Meeting"]. Coarser than events.

Restrict to paid traffic only.

Whose visits credited conversions and revenue are spread over. person: the converting person's own visits (plus a deal's buying committee). company: also every colleague at their company, so a teammate who found the site first shares the credit. Each outcome still counts once. Omit for the workspace setting. Pair with the "companies" metric to count companies instead of people.

Add each metric's previous-period value and change, and the previous totals.

Rank rows by this metric, largest first. Matters only when more rows exist than `limit`.

Rows to keep (series, with granularity). Default and max: as many as fit 200,000 rows.

With by=["ad"], also keep each ad's thumbnail URL, headline, landing page and format.

Instead of the flat arguments, a whole engine query (any kind): "metrics" { window: { period } | { from, to }, metrics: [{ id, as?, events?, types?, models? } | { formula: "demos / visitors", as }], groupBy?, granularity?, compare?: { to: "previous_period" | "previous_year" }, filters?: "dim:op:value" strings or { and|or: [...] }, outcome?: { name, values?, grain? }, sort?, limit? }; "products" { window, compare? } (product sales, refunds, margin); "statement" { window, granularity? } (the P&L); "cohorts" { window, grain?, periods?, groupBy?, spend? }; "paths" { window, step?, outcome? }. Use it for formulas, per-metric events or models, nested filters, previous_year comparisons, products and the P&L.

Rows shown inline (default 50 for metrics, 30 for records, at most 500). The full result is saved either way.

Website domain, e.g. "acme.com". Omit when the account has only one.

Period in plain language: "last 30 days", "yesterday", "July 2026", or "2026-07-01..2026-07-31". Defaults to the last 30 days. For ad_changes, ad_search_terms, ad_pacing (default last 7 days), contacts, deals.

ad_entities, ad_changes: which kinds of entity.

ad_entities: platform statuses, e.g. ["ENABLED"]. deals: open, won or lost.

ad_changes: who made the change.

Ad kinds: one campaign (platform id) and what sits under it.

Ad kinds: e.g. google_ads, meta_ads, microsoft_ads.

ad_entities: settings as they stood at the end of this day (YYYY-MM-DD).

ad_entities: name or keyword contains; companies: name or domain contains; deals: name contains.

ad_search_terms: spent at least this much (workspace currency).

ad_search_terms: only terms with no platform conversion.

contacts, companies, deals: first-touch channel.

contacts: first-touch campaign.

contacts: conversion category.

contacts: exact conversion event.

contacts: needs pii:read.

contacts: needs pii:read.

contacts: sales value at least; deals: amount at least.

deals: one company's deals.

payments: provider event types to keep.

events: pageview, click, form_submit, custom...

sessions: desktop, mobile, tablet.

sessions, events: one visitor (device).

events: only events that are conversions.

Rows at most. Default and max 200,000.

Rows shown inline (default 50 for metrics, 30 for records, at most 500). The full result is saved either way.

What to look for; case-insensitive, matched anywhere in the name.

Where to look. Default: every ad entity kind, plus companies and deals the connection may read.

Ad entities only, e.g. ["ENABLED"].

Website domain, e.g. "acme.com". Omit when the account has only one.

Rows shown inline (default 50 for metrics, 30 for records, at most 500). The full result is saved either way.

Website domain, e.g. "acme.com". Omit when the account has only one.

Period in plain language: "last 30 days", "yesterday", "July 2026", or "2026-07-01..2026-07-31". Defaults to the last 30 days.

Start day (YYYY-MM-DD), instead of period.

Last day, inclusive.

funnel: which saved funnel.

funnel: also count the window before.

stages: keys in order, e.g. ["lead","mql","customer"].

paths: what one step is. Default channel.

paths: journeys behind outcomes (default) or every visitor's journey.

What counts as an outcome. Defaults to `conversions`. ONE request counts ONE kind: 2 MQLs is not a slice of 39 conversions, it is a different object, so never blend them or present a total across them.

• conversions (All conversions, from tracking): Every conversion your tracking script recorded: form submissions, meetings booked, chats, payments. One total, no split.
• conversion_type (Conversions by type, from tracking): The same conversions, split by what kind they were: form fills, signups, meetings, chats, payments and the rest. Which kinds appear depends on what this workspace records. These add up to the conversions total.
• contact_stage (Contacts by stage, from crm): People who reached each stage of your funnel, from your CRM. Counts PEOPLE, not conversions, so these do not add up to the conversions total. Someone now at Customer still counts once under every stage they passed through.
• deal_status (Deals by status, from crm): Deals from your CRM, with the money on them. Counts DEALS, not conversions. Pipeline is the amount on deals created; Won revenue is the amount on deals closed won.
• subscription_status (Subscriptions by status, from payments): Trials, upgrades to paid, and cancellations from your payment provider, with the recurring revenue each one added or removed. Counts subscriptions, not conversions.
• payment_revenue (Payments, from payments): Individual payments you actually received, through Stripe, Shopify or whichever provider you connected. Subscription renewals are not counted here.
• revenue (Revenue, from payments): Every money movement we hold for this workspace: payments and refunds from your payment provider, won deals from your CRM, values you typed on a contact, and anything your tracking script reported. Deduplicated so the same money is never counted twice, and credited to channels by the selected attribution model.

Call get_context first to learn which of these this workspace has data for, and the stage keys to pass as outcome_values.

Narrow the outcome to specific stages, e.g. ["mql"] or ["deal_won"]. Keys come from get_context. Omit to count every value of that kind, which is what "show me our funnel" means.

WHO the outcome is counted over. Only contact_stage offers companies. The company view is SMALLER than the contact view rather than a slice of it: a company sits at the stage of its most advanced contact, so three people at one account are one company. Use it when the question is about accounts rather than people.

Only count these conversion EVENT names, e.g. ["demo_booked"]. Takes precedence over types.

paths: steps kept before a path is cut with "...". Default 5.

paths: paths to keep. Default 1000.

Rows shown inline (default 50 for metrics, 30 for records, at most 500). The full result is saved either way.

Website domain, e.g. "acme.com". Omit when the account has only one.

Which customers: those acquired (first paid) in this period. Defaults to the last 30 days; cohorts usually want 6 to 12 months.

Split by acquisition channel, source, medium, campaign (ltv, curve, revenue_cohorts) or ad_platform (revenue_cohorts).

curve and revenue_cohorts: the age unit. Default month.

revenue_cohorts: ages to return, default 12, at most 60.

revenue_cohorts: add cohort_spend (marketing cost of the cohort's period) for payback.

ltv and curve: the figures. Defaults to lifetime value, customers and average LTV (curve: revenue per customer).

Rows shown inline (default 50 for metrics, 30 for records, at most 500). The full result is saved either way.

From get_records kind="contacts" (id) or search.

From get_records kind="companies" or search. Merges everyone at the company.

From get_records kind="deals" or search. Merges everyone on the deal.

How far back to look. Default 365.

Maximum rows from the event store. Default 2000.

One SELECT (or WITH ... SELECT). Tables are result ids.

Rows inline, default 100, at most 500. The whole output is saved as a result.

A result id (r_ followed by 16 letters and digits) from a prepare tool, query_result or list_results.

Columns to return. Default all.

Sort by this column.

Rows to skip. Default 0.

Rows to return. Default 100, max 2000.

A result id (r_ followed by 16 letters and digits) from a prepare tool, query_result or list_results.

Default csv.

Name of the downloaded file, e.g. "campaigns-2026-q3".

Default 20, max 100.

Website domain, e.g. "acme.com". Omit when the account has only one.

The doc id from list_docs.

Website domain, e.g. "acme.com". Omit when the account has only one.

A template key, e.g. client_report, ad_scaling_plan, traffic_dashboard.

The whole doc as Markdown. Use instead of a template.

Up to 50 block operations, applied in order.

ISO date after which the link stops working.

The period the link always shows, e.g. "previous month" (moves forward each month) or "2026-10-01..2026-10-31". Omit to follow the doc's own date range.

Whether viewers may change the date range. Defaults to false with fixed_window, true without.

What the shared page is called for its viewers, e.g. "Acme, October report". Defaults to the doc title.

At least 6 characters. Required for show_contact_details.

Show full emails and phone numbers in lists of people. Needs a password. Off by default.

Websites the link may show (a switcher when several). Defaults to the doc's website.

Older form of allow_date_change: every control viewers may change.

Website domain. Only needed when the credential covers more than one.

Website domain. Only needed when the credential covers more than one.

What the alert is called, e.g. "New demo booked".

The template it started from, e.g. "new_lead".

The alert spec: trigger, filters, gather, condition, compose, deliver, pacing. Take it from a template and edit it.

From list_alerts.

Mark the message as a test. Default true.

The contact id: the id column of get_records kind="contacts", or a contact hit from search.

Lead status, the same field as lead_status. Must be a status this workspace defines; an unknown name is refused with the allowed list.

Lifecycle stage name from this workspace, e.g. "Customer". Unknown names are refused with the allowed list. Propagates to the connected CRM.

Lead status name from this workspace. Unknown names are refused with the allowed list. Propagates to the connected CRM.

Lead quality: good, not_a_fit or spam. Spam leaves every report and is never sent to a CRM or ad platform. "pending" hands the lead back to the workspace's Lead Rules.

Deprecated: use quality. true means good, false means not_a_fit.

What the deal was actually worth.

Deprecated: use quality "spam".

Values for the workspace's custom fields (Service, State...), by field key, e.g. {"service": "AC repair"}. null clears one. A value set here is never replaced by a form answer or a lead rule. Unknown keys are refused with the list of fields.

The company id: the id column of get_records kind="companies", or a company hit from search.

A stage name from this workspace, e.g. "Customer". Pins the company: stage stops being derived.