The stack.
All of it.Five layers. Every block engineered for agents that need to be trusted, tracked, and fast.
Three axioms. No exceptions.
Identity is infrastructure
Every byte that enters Zaps is stamped with a W3C DID. The identity layer sits below the storage layer — there is no code path to unattributed data. This is not an ACL grafted onto an existing system. It was the first thing we built.
Purpose over configuration
You do not configure a Zaps database. You declare its purpose. The platform derives TTL policy, eviction strategy, durability mode, engine selection, and indexing behavior from a single enum variant. Seven purposes, zero knobs.
Engines are swappable
The KvEngine trait decouples storage from protocol. Zaps ships on AkashaKV by default, with all purposes isolated by keyspace over the same durable engine. You can add adapters without changing the protocol surface.
The full picture.
Zero-Trust Identity
Every operation authenticated via W3C DID. No anonymous access path exists.
Purpose Databases
Seven profiles — cache, session, agent-memory, rate-limiter, queue, feature-flag, audit-log.
Low-Latency Durable KV
AkashaKV dual-mode keeps hot reads fast while the LSM tree keeps data durable.
Adapter-Friendly Engine Layer
Storage remains behind KvEngine, so future adapters do not change HTTP, RESP3, or WebSocket clients.
Full Audit Trail
Hash-chained, tamper-evident log of every mutation. BLAKE3 fingerprints. Exportable.
Real-Time Subs
WebSocket key-change subscriptions scoped per principal. Pattern matching.
Redis Compatible
RESP3 wire protocol. redis-cli, redis-py, ioredis — your existing tools work.
Per-Principal Metering
Usage tracking and quota enforcement at the identity level. Per-agent cost attribution.
Durable Storage
Akasha LSM-tree backend with write-ahead log, MVCC, Snappy compression, point-in-time recovery.
Tenant Isolation
Keyspace isolation enforced at the protocol layer. Agents cannot read other agents' keys.
Multi-Protocol
HTTP REST, WebSocket, and RESP3 on a single deployment. One port, three wire formats.
Full Observability
OpenTelemetry distributed traces, Prometheus metrics, structured JSON logs. Built in, not bolted on.
Ready to explore?
Start with purpose databases, dive into the architecture, or jump straight to the quickstart.