SpiceDB is an open-source, Google Zanzibar-inspired database designed for storing and computing permissions data at scale. It enables developers to build fine-grained authorization services by modeling relationships between objects and computing permissions from those relationships. SpiceDB uses a custom schema language called SpiceDB Schema to define object types, relationships, and permissions. This schema supports CI/CD integrations for validation and testing, and includes a built-in testing server for running API calls against test data. The system exposes both gRPC and HTTP/JSON APIs, allowing client applications to perform authorization checks. It uses a distributed, parallel graph engine based on the architecture described in Google's Zanzibar paper, and implements a flexible per-request consistency model that addresses the "New Enemy Problem" — a class of authorization bug where users may incorrectly gain or retain access. SpiceDB supports multiple pluggable storage backends including in-memory, Google Spanner, CockroachDB, PostgreSQL, and MySQL. For observability, it integrates with Prometheus for metrics, pprof for profiling, structured logging, and OpenTelemetry for distributed tracing. A Kubernetes Operator (spicedb-operator) is available for managing SpiceDB clusters, supporting zero-downtime datastore migrations and automated update channels. SpiceDB is released as open source software and is productized by AuthZed as a managed cloud service. Notable production users include Reddit, Matillion, and Netflix.