Why Zaps
Redis was designed for caches. Zaps was designed for agents. The difference shows up the moment you give a thousand independent actors shared write access — Redis becomes a debugging nightmare; Zaps stays auditable, scoped, and predictable.
Pick a language, copy the install command, follow the quickstart. Need depth? Concepts explain the architecture; guides cover the patterns enterprise teams reach for; the API reference is generated from a single OpenAPI spec.
Redis was designed for caches. Zaps was designed for agents. The difference shows up the moment you give a thousand independent actors shared write access — Redis becomes a debugging nightmare; Zaps stays auditable, scoped, and predictable.
Five layers — Protocol, Edge, Engine, Persistence, Cross-cutting. One port handles RESP3, gRPC, and WebSocket. AkashaKV dual-mode storage powers the production path, with future adapters available behind the KvEngine trait.
Every operation arrives with a verified W3C DID. Authorization runs at the edge, before the engine sees the request. There is no anonymous code path. API keys, OAuth 2.0 client credentials, JWT bearer, mTLS — all four are first-class.
Seven workload-shaped profiles: agent-memory,cache, session,rate-limiter, queue,feature-flag, audit-log. Each ships engine, eviction, TTL, and replication defaults calibrated to its workload. You name what you're storing; Zaps picks the configuration.
Every write is captured in a tamper-evident BLAKE3 chain before it commits. The audit log is exportable as a single signed file with cryptographic proof of completeness.
WebSocket key-change subscriptions scoped per principal. Pattern matching with glob support. Sub-millisecond fan-out at 10,000 subscribers.
Keyspace isolation is enforced at the protocol layer, not in application code. Agents cannot read other agents' keys — verified by the audit chain.
99.99% on Bolt and above. Engine hot-swap covers most outage classes. Crash recovery in 50–500ms via WAL replay on top of the latest snapshot.
These pages are served by Next.js right now. The full Mintlify project lives in /zaps/mintlify-docs and is configured in mint.json with the Zaps brand already applied.
When the production Mintlify origin is configured (set theZAPS_DOCS_HOST environment variable and rebuild),/docs/* traffic on this domain proxies through to Mintlify automatically. URL slugs are kept identical between this page and mint.json so links keep working.
Owners by section are tracked in mintlify-docs/README.md. Quarterly audits against the SDK & Documentation mandate (April 2026) are the responsibility of those owners.
Spot something missing or wrong? File an issue at github.com/l1feai/zaps/issues — triaged within one business day.