Rawbbit: self-hosted event tracking and analytics pipeline
Open-source game telemetry pipeline: you keep the raw data, agents query it over MCP.
Rawbbit (formerly DataQueryEvent) is a self-hosted event telemetry pipeline for mobile and web games. It helps small teams collect player events, sessions, funnels, retention signals, economy and monetization events while keeping ownership of the raw data, without making a proprietary vendor the center of their telemetry stack.
The main path:
- Game clients, web builds, mobile apps and backend services send batched events over HTTP to a collector API that validates and enriches them.
- NATS JetStream buffers events, so request handling is separated from storage writes.
- A raw writer lands partitioned Parquet files in S3-compatible storage (SeaweedFS). Raw Parquet is the system of record.
- Scheduled dbt Core runs load bounded Parquet windows into ClickHouse, the main analytics database.
- Data is served through an MCP server for AI agents (Claude, Codex, OpenCode, OpenClaw), Metabase dashboards and direct SQL.
Deployment is split across two VMs: one for ingestion and raw storage, one for analytics and access, with Ansible automation and Dozzle for logs. An optional path loads the same Parquet into BigQuery via external tables and SQLMesh.
It also includes Unity and browser-game SDKs, plus coding-agent plugins for Claude and Codex that propose an event map for a game and integrate the approved events.
Need something similar? mirlan@dataengineer.fi
Next: ROAS pipeline for a mobile game →