porcupineScope and roadmap#
Available in v0.10.0#
- Capture. Send events through the native or Segment interfaces, or point a supported PostHog SDK profile at Porcupine. Stable producer IDs support retries. Historical PostHog events and persons can be imported from JSONL.
- Analytics. Discover event names, observed properties and the latest example UUID. Run read-only ClickHouse SQL with typed parameters. The query reference describes the supported SQL and the 60-second cache. Identified capture links anonymous activity to a user. Cookieless capture estimates daily visitors.
- Delivery. Export a query once, send events to a webhook individually or in batches, or schedule a query pipeline to publish time windows to S3.
- Management. Manage projects, including rename and deletion, through HTTP, the CLI or MCP. Humans can authorize tools using email verification or configured OIDC. Service accounts can use scoped credentials.
- Operations. Inspect the instance status through HTTP, CLI, or MCP. The single-host deployment uses bounded durable queues, optional cold storage, and a backup and restore procedure.
Areas for future work#
Agents currently write the SQL for each question. Named metrics, a saved and parameterized query catalog, built-in session definitions and live-presence rules are candidates for future work. Funnels and retention can already be expressed in the supported SQL dialect when the events carry the identities and timestamps the analysis needs.
A replicated queue and multiple ingestion workers would change the failure model and deployment requirements. The current architecture targets a single application process with local durable storage.
The compatibility matrix records the supported PostHog SDK behaviors. General aliasing, merging two identified users, unlinking identities, feature flags and replay are outside that contract. The query API uses ClickHouse SQL.