Returns merged PR/MR counts grouped by platform login for the organization, filtered by VCS source (github or gitlab). With granularity=day, returns one row per (login, calendar day) instead of a single per-login total; days are UTC calendar dates (the underlying metric_date has no time component) and are zero-filled for logins with at least one merged PR in range, so quiet days are returned as 0 rather than omitted. Days older than the trailing ~30-day sync window are frozen at their last sync and may miss late-recorded activity; day-level history exists from the org's first sync onward. granularity=day is capped at 35 days per request; use granularity=window for longer ranges. granularity=week or month returns one row per (login, bucket) instead, zero-filled per bucket rather than per day; weeks are ISO-8601 (Monday start) and months are calendar months, both in UTC. week and month are capped at 400 days per request. prsMergedWithCodingTool counts merges whose commits matched the tenant's configured co-author patterns (Claude Code's by default), so it answers 'AI-assisted' without naming a vendor; the additive prsMergedByVendor / totalPrsMergedByVendor maps name the vendor for every merge that matched a recognized vendor's patterns. The two are NOT nested: a tenant that configures a signature no shipped vendor default covers gets the merge counted in prsMergedWithCodingTool with no vendor identified. Because one PR can be attributed to several vendors their values may sum to more than prsMerged. Vendor keys are CodingAssistantMetricProvider names (e.g. ClaudeCode, not claude-code) and a key is present only with a non-zero count. The maps are omitted from the payload when no vendor was attributed, and unlike prsMerged they are not zero-filled on week or month buckets with no attributed merge.