We are building the infrastructure for the autonomous enterprise. No black boxes, no vendor lock-in. Our core "Spotlight" tools are open-source and protocol-aware, designed to extend your existing Kafka investments into a secure, reasoning-capable mesh.
We don't build wrappers. We build the foundation. Our open-source spotlight tools, KafScale, KafClaw, KafGraph, kafSIEM and Scalytics Copilot, are protocol-aware and built to extend your existing Kafka investments into a reasoning-capable mesh.
Each of the four kaf* products is independent and runs on any Kafka-compatible infrastructure. Combined, they form the most complete vendor-neutral stack for agentic AI on Kafka.
KafScale specifically is the only Apache 2.0 licensed, S3-native, unlimited-scale Kafka-compatible platform on the market. No BSL restrictions, no vendor control plane, no usage fees. We provide the architecture in the open and the PMC-level expertise to deploy it at scale in your mission-critical environment.
KafScale is a Kafka-compatible streaming layer built natively for S3. It eliminates the "Storage Tax" of traditional brokers by using object storage as the primary source of truth.
The Power:
~200ms latency for durable agent state.
The Benefit: Infinite retention of every agent thought at S3 prices.
KafClaw is the orchestration mesh for multi-agent coordination. Built in Go for maximum efficiency, it manages observable groups and skill routing through typed envelopes.
The Power: Typed Envelopes, Observable Groups, Shared Memory.
Modes: Local, Local-Kafka (Orchestrated), or Remote
KafGraph is the distributed shared memory for agentic systems. Driven by events from Kafka and backed by BadgerDB, every agent sees the same state without a central database.
The Power: Event-driven updates, BadgerDB reads, replayable state.
The Benefit: Infinite memory for agents at log scale.
kafSIEM is a typed-entity graph for link analysis over live event streams. Every edge carries a citation to the source event, giving analysts auditable reasoning instead of opaque queries.
The Power: Typed entities, citation edges, real-time ingest.
The Benefit: Defensible link analysis for regulated environments.
The Copilot is your private model-hosting engine. It integrates with MCP (Model Context Protocol) to reason over your live Kafka topics without data ever leaving your secure execution perimeter.
The Power: Private LLM Hosting, MCP integration, Grounded Context.
The Benefit: Enterprise-grade management for regulated industries.
Wayang is an Apache Top-Level Project for cross-platform query processing. Scalytics was founded by Wayang's original inventors and maintains core capabilities used at Fortune 500 scale.
The Power: Cross-platform planning across Spark, Flink, and Databases.
The Benefit: Compute moves to where the data lives.
Command-line network sniffer for Apache Kafka. Tests L3-L7 layers (DNS, TCP, TLS/mTLS, SASL, protocol, produce/consume) in one tool. Agent-optimized with JSON/HTML reports.
GitHub: https://github.com/scalytics/kshark-core
High-performance mirroring between Kafka clusters using franz-go. Includes real-time dashboard, AI anomaly detection, incident insights, compliance reports, and recovery tools.
As Apache Wayang PMC members and KafScale maintainers, Scalytics engineers contribute to core infrastructure projects used by Fortune 500 companies. Our work focuses on Kafka scaling, federated analytics, and private AI ops, delivering production-proven tools for enterprise streaming.
"Scalytics contributions have accelerated Wayang's hybrid batch/stream capabilities, enabling multi-runtime analytics at scale." - Apache Wayang
Our contributions power resilient streaming for enterprises. 100+ stars, active issues, merged PRs across projects.
Our tools are open because the enterprise deserves transparency. But building a resilient, auditable agent mesh on legacy infra is a high-stakes engineering feat. We live from our consulting - helping you bridge the gap between a "Local" assistant and a global "Agentic Message Bus."
