Frequently asked questions

Zabta is an AI agent governance platform built by Zainova Labs LLC. It helps businesses govern autonomous AI agents through real-time policy enforcement, regulatory compliance scoring, and trust monitoring. The platform evaluates every agent action against a three-layer policy engine covering 14 jurisdictions, providing compliance dashboards and risk scoring so teams can deploy AI agents responsibly under frameworks like the EU AI Act and Colorado AI Act.

Zabta is designed for tech leads and engineering teams at startups with 10 to 50 people who are deploying autonomous AI agents and need compliance infrastructure but don't know where to start. It is also useful for any organization building AI agents with frameworks like OpenAI, Anthropic, LangChain, CrewAI, or AutoGen that needs to ensure regulatory compliance before enforcement deadlines.

No. Zabta is not an AI agent framework and does not build, host, or run agents. It is a governance layer that sits alongside your existing agent infrastructure to enforce policies, score compliance, and monitor trust. You continue using your preferred agent framework (OpenAI, LangChain, CrewAI, etc.) and add Zabta as the compliance and governance layer.

Zabta uses a three-layer policy system. Universal policies apply to all agents everywhere and cover fundamental safety rules. Jurisdiction Packs add region-specific regulatory requirements for each of the 14 supported jurisdictions, such as EU AI Act rules or Colorado AI Act requirements. Sectoral Overlays add industry-specific rules for domains like healthcare, finance, and education. Every agent action is evaluated against all applicable layers in real time, returning a decision of allow, deny, or escalate along with the specific policy, citation, risk tier, and evaluation time.

Zabta covers 14 jurisdictions with real regulatory content mapped to enforceable policies. Key regulations include the EU AI Act (enforcement begins August 2026), the Colorado AI Act (effective June 2026), GDPR, CCPA, and jurisdiction-specific AI governance frameworks. Each jurisdiction pack contains actual regulatory requirements, not placeholder metadata, ensuring compliance scoring reflects real legal obligations.

Zabta provides three purpose-built dashboards. The Compliance Dashboard shows real-time compliance scores per agent and per jurisdiction, with risk tier breakdowns and policy violation details. The Trust Dashboard monitors agent behavioral consistency, escalation patterns, and trust score trends over time. The Health and SRE Dashboard tracks operational metrics including evaluation latency, error rates, uptime, and system health for your agent governance infrastructure.

Install the Python SDK with pip install zabta (version 0.3.0 on PyPI). Configure your API key (keys use the zbt_ prefix) and call the evaluation endpoint before or after each agent action. The SDK provides auto-instrumentation for OpenAI, Anthropic, and LangChain, meaning you can add governance with minimal code changes. The SDK returns structured responses including decision (allow, deny, or escalate), reason, policy citation, risk tier, and evaluation time in milliseconds.

Zabta is framework-agnostic and works with any AI agent implementation. The Python SDK provides built-in auto-instrumentation for OpenAI, Anthropic, and LangChain. Agents built with CrewAI, AutoGen, or custom frameworks can integrate through the SDK's evaluation API or the REST API directly. Zabta governs agent actions regardless of which underlying model or framework powers the agent.

Free covers cloud policy evaluation with up to 2 jurisdiction packs, 3 agents, 1 seat, and 7-day audit history — no card required. Starter is $99/month and unlocks all 14 jurisdiction packs and all 5 sectoral overlays, Broker credential governance, the compliance dashboard, 25 agents, 3 seats, and 90-day audit history — regulatory coverage isn't metered by count on paid plans. Pro is $399/month and adds PDF compliance and evidence export, unlimited agents, 10 seats, 1-year audit history, and priority support. There's also a Design partner track for teams deploying agents now who want early access and a direct line to the founders — email hello@zabta.ai. There is no per-evaluation metering on any plan.

Zabta maps EU AI Act requirements to enforceable policies within its jurisdiction pack system. This includes risk classification of AI systems, transparency obligations, human oversight mandates, and documentation requirements. The compliance dashboard provides a real-time score showing how each agent measures against EU AI Act requirements, identifying gaps before the August 2026 enforcement deadline. The platform evaluates every agent action against these mapped policies and flags violations with specific regulatory citations.