Immutable audit trail, real-time policy enforcement, and EU AI Act compliance — for every AI agent action. In 3 lines of code.
Three steps from zero visibility to full governance. No infrastructure changes. No agent rewrites.
One pip install. Works with LangChain, CrewAI, OpenAI Assistants, or any custom Python agent. No agent rewrites needed.
pip install horkos-sdkEvery action evaluated before execution. Block SQL injections, prompt injections, require approval above risk thresholds — all configurable as JSON. Zero code needed.
< 2ms overheadEU AI Act, DORA, and SOC2 reports generated automatically from your audit trail. One click for your auditors. Compliance score. Gap analysis. Recommendations.
PDF exportEverything you need to govern AI agents at scale. Nothing you don't.
JSON rules evaluated in real-time. Block injections, require approval above risk thresholds, restrict action types per agent. Runs in <2ms with zero infrastructure changes.
Every action logged permanently — input, output, timestamp, user, duration, cost. Cryptographically verifiable. Cannot be modified or deleted, even by admins.
High-risk actions pause and notify approvers via Slack. Full context delivered. Approve or deny in one click. Agent resumes or stops based on decision. Automatic timeout.
Auto-generated reports for EU AI Act, DORA, and SOC2. Compliance score, gap analysis, findings, and actionable recommendations — ready for your auditors in one click.
AI-powered detection of unusual agent behavior — abnormal action volumes, unexpected access patterns, cost spikes, and risk escalations. Instant alerts.
Deploy Horkos in your own infrastructure. Kubernetes-ready Helm chart. Zero data leaves your environment. GDPR-compliant by design. Full EU data sovereignty.
Add Horkos in 5 minutes. Free forever for the first 1,000 actions. No credit card needed.