Skip to main content
The session-start tool. Call this once before any analytics question so the agent knows the user’s timezone, currency, configured custom event labels, connected ad accounts, and default attribution model / window. Replaces the legacy “data instructions” prompt-stuffing pattern with a single, cacheable, ~1 KB call.

When to use

  • At the start of every chat session before any analytics question.
  • Whenever the user references a custom event by name (“deals”, “lead magnet”) and you need the slot map.
  • To check the default attribution model before running query_ad_metrics.
For per-question lookups about a specific entity, use search_workspace. For tracking-health diagnostics, use get_space_health.

Parameters

Output shape

ai_credit_balance reflects the team’s ai_credits column (shared across every space on the team). default_attribution_model / default_attribution_window surface Cometly’s product defaults — they aren’t currently persisted per-space. business is null until setup step 1 (“About your business”) has been saved for the space. summary is up to three sentences on what the company sells, to whom, and how a customer converts on the site; it is drafted automatically from the website and can be edited in Setup → step 1. type is one of ecommerce, saas, lead_gen, services, or other. lead_destination is where leads or sales are recorded (for example HubSpot or Shopify checkout), uses_native_lead_forms covers Meta, LinkedIn, and TikTok native lead forms, and ad_platforms lists the platform keys the user says they run.

Caveats

  • The ai_credit_balance field reflects the team-level balance at request time. Don’t recommend analyze_report (or other credit-charging tools) when the balance is at zero.
  • last_synced_at is best-effort — different ad platforms expose different timestamp columns. Use get_space_health for the canonical freshness check.