MCP Tools
The v1 tool list, their scopes and their cost
Twelve read-only tools. Weight is what a call costs against the per-minute limit (120 points per token, 600 per organization) and the monthly quota.
| Tool | Scope | Weight | Returns |
|---|---|---|---|
list_projects | projects:read | 1 | projects of the organization the token can see |
list_event_names | projects:read | 1 | event names with their property keys (the schema) |
get_event_property_values | projects:read | 1 | most common values of one property |
get_timeseries | analytics:read | 1 | users and events per day |
get_top_events | analytics:read | 1 | most frequent events in a range |
get_overview | analytics:read | 5 | headline KPIs (users, events, visits, page views, bounce, duration) |
query_events | analytics:read | 5 | raw events with filters, capped to 200 rows, contact details masked |
funnel | analytics:read | 5 | ordered conversion, 2 to 10 steps |
retention | analytics:read | 5 | weekly cohorts |
segment_by_property | analytics:read | 5 | one event broken down by a property |
list_users | users:read (+ users:pii to unmask) | 5 | tracked users, capped to 100 rows |
export_events_sample | export:read | 5 | a CSV sample, 1,000 rows at most |
Inputs
Every analytics tool takes project_id, start and end (ISO 8601, default last 30 days). Funnels take steps and window_hours; retention takes periods (1 to 12); segments take event_name and property_key.
Outputs
Tools return structured JSON matching a published output schema plus a one-paragraph text summary, so both agents and people can read them. Errors come back as tool errors with a message (Insufficient scope: users:read required, monthly quota reached).
Resources and prompts
The server also publishes the project schema and these docs as resources, and four prompts: weekly_report, funnel_analysis, retention_review and anomaly_check.
Privacy
query_events and list_users mask emails and phone numbers unless the token has users:pii. Exports are samples, not full dumps; use the dashboard export for complete data.