ZIZKA AI S.L.
ZIZKA AI builds operational infrastructure that helps teams make AI agents reliable in production. Its flagship ZizkaDB records agent decisions and execution so teams can replay sessions, trace causal lineage, detect behavioral drift, and support auditability and EU AI Act-oriented oversight.
At a Glance
- Teams and startups shipping AI agents into production
- Enterprises operating multi-agent fleets
- Regulated industries and organizations preparing for EU AI Act obligations
- Engineering, security, and compliance teams needing private deployment and audit evidence
- +1 more
AI Tools by ZIZKA AI S.L.
(1)
ZizkaDB
AI Agent Audit Database
Discussions
No discussions yet
Be the first to start a discussion about ZIZKA AI S.L.
Latest News
Published Understanding Agent Behavior Drift: A Pre-Diagnostic Case Study, describing ZizkaDB behavior baselines, drift scoring, event/transition deltas, and pre-diagnostic investigation.
Published How to Make Your AI Agent EU AI Act Compliant: A Practical Guide, showing how ZizkaDB provides runtime audit-trail evidence.
Published The Science of Machine Learning vs. the Push for AI Deployment, arguing for auditable AI with human oversight rather than assumed deterministic behavior.
Published Cloud Hosting Is the Data Hostage Model, and It Has No Place in the AI Era, advocating user-controlled/self-hosted infrastructure.
Products & Services
AGPL-3.0 open-source operational database and audit trail for AI agents, deployable with Docker Compose on a customer's infrastructure. Includes API, tenant dashboard, SDKs, MCP server, activity/behavior/reports/suggestions views, session replay, causal lineage, drift detection, semantic search, point-in-time state, and GDPR erasure capabilities.
Hosted ZizkaDB at db.zizka.ai with zero-operations Postgres/vectors/dashboard, API keys, session replay, causal lineage, behavioral drift monitoring, and Fleet ranking across agents and projects.
Single-tenant private-cloud deployment adding fleet dashboard and ranking, audit export with checksums, commercial licensing, supported installation, integration workshops, and data residency in the customer's VPC.
Market Position
ZizkaDB positions itself as an agent-operations database and system of record focused on decision history, causal lineage, replay, and behavioral drift—not a RAG vector database, transactional application database, or conventional distributed-tracing product. Its differentiation is the same open-core engine across self-hosted, managed cloud, and Enterprise VPC, with framework-agnostic integrations. The Enterprise page explicitly contrasts its agent-ops layer with vector databases such as Pinecone/Qdrant and tracing tools such as LangSmith/OpenTelemetry.
Leadership
Founders
Mir Arshad Talpur
Founder of ZIZKA AI S.L.; a serial founder with a background spanning software engineering, artificial intelligence, and systems thinking. He describes himself as an NVIDIA-certified Agentic AI professional with experience delivering software across US, EU, and Gulf markets.
Executive Team
Mir Arshad Talpur
Founder
Serial founder and NVIDIA-certified Agentic AI professional with experience in software engineering, AI, and systems delivery across US, EU, and Gulf markets.
Saad Amjad
Founding Engineer
Full-stack engineer with seven years shipping mobile and web products to millions of users. Previously a core engineer at Retailo, built the Hao wellness app solo from MVP to App Store launch, and worked at Washmen across React Native, partner PWAs, Sails.js, and AWS.
Board of Directors
Founding Story
Founder Mir Arshad Talpur says ZizkaDB came from observing that teams shipping agents into production need operational data—replay, lineage, and drift detection—not another thin language-model wrapper. The company chose to build a purpose-built, predictable and auditable store for agent behavior; its name references Jan Žižka's use of strategy and technology to give a small force leverage against larger armies.
Business Model
Revenue Model
Open-source AGPL self-hosting is free; revenue comes from managed-cloud subscriptions and annual Enterprise VPC commercial licenses, installation, workshops, and support.
Pricing Tiers
AGPL open-source stack running on the customer's infrastructure.
Managed cloud plan with 50,000 events/month and 2 API keys.
Managed cloud plan with 100,000 events/month and 5 API keys.
Single-tenant VPC, up to 5 agents, Fleet dashboard, audit export, commercial license, and installation.
Single-tenant VPC, up to 50 agents, Fleet dashboard/ranking, checksum audit export, commercial license, installation, and integration workshop.
Single-tenant VPC with unlimited agents, Fleet dashboard/ranking, checksum audit export, commercial license, and dedicated install sprint.
Target Markets
- Teams and startups shipping AI agents into production
- Enterprises operating multi-agent fleets
- Regulated industries and organizations preparing for EU AI Act obligations
- Engineering, security, and compliance teams needing private deployment and audit evidence
- Developers using Python, TypeScript, LangChain, CrewAI, LiveKit, MCP, or REST
- Debugging production AI-agent incidents and finding root causes
- Monitoring prompt, model, tool, and user-driven behavioral drift
- Audit trails and compliance evidence for regulated or high-risk AI systems
- Operating and ranking fleets of multiple agents
- Customer-support and business-process agents
- Voice agents and agents built with LangChain, CrewAI, LiveKit, MCP, or custom REST clients