Porcupine documentation#

Porcupine collects product events and makes them queryable through ClickHouse SQL. Connect an agent through MCP, use the standalone CLI, or call the HTTP API. Your agent generates the analysis and charts. Porcupine has no analytics dashboard.

Start here#

If you are trying Porcupine for the first time, follow the local quickstart. It starts an instance, sends a test event, and queries it.

If your team already has an instance, connect your tools, then query your events. You need the instance URL and access from its administrator.

To let an agent guide you, give it the agent guide and tell it whether you want to deploy an instance or connect to an existing one. Every guide also has a View Markdown link.

Collect and use data#

Task	Guide
Send events from a browser or backend	Choose an input
Keep using a PostHog SDK	Configure PostHog capture
Link anonymous activity to a signed-in user	Users, identity, and sessions
Bring your existing PostHog history	Import files with resume
Discover events and answer a question	Query your events
Download a large query result	Export query results
Deliver events or scheduled query results elsewhere	Webhooks and S3 pipelines
Fix or remove stored events	Event correction and deletion

Run an instance#

Start with hosting. Then set up team sign-in, manage projects, and invite people. For a backend, cron job, or shared agent, create a service key.

The deployment operator configures S3 cold storage and backups. These settings apply to the deployment and are separate from project permissions. You can also read an authenticated instance status report.

Look up a contract#

Use the HTTP API reference, SQL reference, or MCP reference for request fields, permissions, and limits. The running instance serves its OpenAPI contract at /openapi.json. In MCP, porcupine_docs describes the operations available to your credential.

For deployment constraints and design decisions, read the architecture, durability contract, and current scope.