What ZizkaDB is
ZizkaDB stores agent decisions, tool calls, and outcomes with causal links. It is not a vector index for RAG documents and not distributed traces. Most teams use all three alongside each other.
Works with any LLM framework. No framework lock-in. Instrument with Python, TypeScript, MCP, or REST. Open core engine on GitHub (AGPL) runs the same code on self-host, managed cloud, and Enterprise VPC.
Multi-agent fleets
Each agent logs under a distinct name. Tenant-wide API keys support SaaS patterns with thousands of logical agents.
See all agents, last seen, and session volume in one place. Included in your Enterprise VPC deployment.
Compare recent sessions to each agent's baseline. Verdicts from Stable to Significant. Operational behavior change, not hallucination detection.
Walk decision chains with why() and replay sessions when something breaks in production.
Capabilities
The same ZizkaDB engine runs on GitHub (AGPL), managed cloud, and Enterprise VPC. Enterprise adds commercial license, fleet UI, audit export, and supported install.
Why enterprises choose ZizkaDB
Built for teams that cannot send agent logs to a shared cloud and need a complete operational record for every agent decision.
Security and trust
Single-tenant deployment with no shared infrastructure. Your data never leaves your network.
Deployment
Same install package every customer. Only variables: region, sizing, DNS, secrets, and license tier.
Add ZIZKADB_HOST and ZIZKADB_API_KEY to your services. Instrument five log points per conversation.
You merge and redeploy through your pipeline. Optional Launch Sprint can help instrument one or two services via pull requests.
Agent subnets reach ZizkaDB via same VPC, VPC peering, or PrivateLink.
Pricing
Contact us for exact pricing. All tiers are annual licenses with commercial rights included.
FAQ
Get started
Tell us about your agent fleet. We will map a 4 week install path or point you to managed cloud.
Technical resources