/v1/analytics. Use them to build dashboards, monitor quality and cost, and spot regressions.
All analytics endpoints are GET, org-scoped, and require the analytics:read permission. Most support a time range and return either a timeseries, a breakdown, or a rolled-up dashboard summary.
Analytics domains
Sessions
/v1/analytics/sessions — volume, channel mix, resolution and first-response time, escalation rate, turn depth.Tools
/v1/analytics/tools — invocation counts, latency, error and retry rates, reliability rankings per tool.Voice
/v1/analytics/voice — call volume, answer rate, outcomes, and warm-transfer performance.Knowledge
/v1/analytics/knowledge — retrieval volume, top documents, ingestion and freshness metrics.Model router
/v1/analytics/model-router — cost by model, provider, assistant, and operation; latency, TTFT, fallbacks, and cache savings.Policies
/v1/analytics/policies — violation rates, action breakdowns, and monitor-mode rollout tracking.Integrations & mocks
/v1/analytics/integrations and /v1/analytics/mocks — connection health, invocation success, and mock match rates.Memory
/v1/analytics/memory — memory coverage and ingestion over time.Knowledge gaps
Beyond usage metrics, Feather clusters unanswered or low-confidence questions into knowledge gaps so you know what content to add. Query them under/v1/knowledge-gaps: a summary, the clusters themselves, per-cluster evidence and lifecycle (owner, status), and document-health signals.
Using the data
The analytics surface is broad — one endpoint per metric and view — so the fastest way to find what you need is the API reference. Each endpoint documents its exact query parameters (time range, bucket granularity, filters) and response shape.Analytics API reference
Browse every analytics endpoint, its filters, and its response schema.
Analytics endpoints are read-only reporting views. To act on data — retry a
tool, re-run an evaluation, purge a conversation — use the corresponding
feature API.