The business data you can query

One data spine, four surfaces

Everything below is available identically to the Assistant, the Insights deep-dives, your [MCP endpoint](/guides/expose-your-business-as-an-mcp-endpoint) and the CLI — because they all call the same tool set over the same first-party data spine. Access is scoped and redacted by the caller's access policy.

Analytics & funnels

First-party, cookieless web analytics: visitors, leads, conversions and revenue in funnel order, plus segment cuts and trends. Tools: get_analytics, get_analytics_segment, get_analytics_trend, get_funnel_analysis, get_attribution, list_goals, list_funnels. Every metric block states its unit (visitors are distinct people, conversions are sessions), every windowed read returns the window it actually covered, comparisons carry the previous window's values, and the funnel is a true cohort (each step is a subset of the previous one). The AI cites the metric, timeframe and source for every number.

Heatmaps & session replay

In-house click/scroll heatmaps and rrweb session replay (Growth and above). Tools: get_heatmaps (most-clicked elements, rage/dead clicks, scroll reach), list_session_recordings, and `analyze_session_recording` — which turns a raw recording into a behavioural narrative the AI can reason over. Ask it to review your most frustrated sessions and it will tell you exactly where people rage-click or abandon.

CRM & identity graph

The contact identity graph stitches anonymous visitor → email → customer. Tools: list_contacts (paged; kind says person vs company; every lifecycle stage incl. cold), search_contacts, get_contact_timeline (a person's full journey), get_visitor_journey (an anonymous visitor's journey), get_lead_scores. Writes — update_contact, add_contact_note, log_contact_activity — execute immediately under the key's policy. send_email never sends directly: it returns status: pending with an approval_id, the workspace owner approves it in the dashboard, and get_approval_status / list_pending_approvals tell you what happened.

Company enrichment

Identify the companies behind anonymous B2B visitors and enrich contacts with firmographics (industry, size, revenue band, location). Tool: get_company_enrichment. Paid-plan, policy-gated on the enrichment data class.

SEO, ads & social

get_seo_insights (Search Console top queries + striking-distance keywords), get_ads_stats (Meta + Google spend/ROAS tied to first-party revenue), get_social_stats (organic follower + engagement stats), plus audience tools (list_audiences, create_audience, sync_audience) to build and push segments to ad platforms.

The company brain

search_company_brain searches a durable knowledge library built from past conversations, SOPs and connected Slack/Gmail/Docs — so the AI recalls prior decisions instead of re-asking. web_research (Tavily) covers anything outside your own data, with citations.

Run the analyst itself

run_analyst queues a full run of the same AI business analyst that powers the dashboard — every tool above, connected tools, web research, approval-gated actions — and returns a thread_id; get_analyst_run polls for the reply and any approvals it created. It counts as one AI run against the plan and runs under the calling key's access policy. The CLI's /api/v1/runs is the same pair over REST.

Example prompts

• "Analyze the checkout funnel and show me the biggest drop-off step with the numbers."
• "Watch our three most recent rage-click sessions and tell me what's broken."
• "Which companies visited the pricing page this week but didn't convert?"
• "Compare this month's conversion rate to last month by channel."

For the endpoint-native versions of these, see the prebuilt MCP prompts.