Kaplaix is a governance platform designed to provide real-time oversight, risk evaluation, and compliance evidence for AI agents operating in production environments. It addresses the challenges organizations face in monitoring autonomous AI agents, helping security, compliance, and engineering teams understand and control agent actions before incidents or regulatory audits occur. The platform aligns with regulatory frameworks such as the EU AI Act (covering Articles 9, 12, 13, 14, and 17) and SOC 2 Trust Services Criteria, offering features tailored to meet these standards.
Key capabilities include real-time risk scoring based on the OWASP Agentic Top 10 dimensions, pre-execution policy gates that can approve, deny, or escalate tool calls, and a circuit breaker mechanism for repeated risky behavior. Kaplaix captures every agent session in detail, enabling causal-chain reconstruction, narrative summaries, and key-moment detection to explain agent decisions. Each session is recorded as a signed, tamper-evident evidence package, with PDF reports, JSON event logs, and SHA-256 manifests, suitable for audit and regulatory review. Exports can be mapped article-by-article to EU AI Act requirements, and the architecture supports tenant isolation, role-based access, immutable audit logs, and scrypt-hashed credentials.
The platform is designed for integration with a wide range of agent frameworks and supports open ingestion via HTTP, SDKs, OTLP, or framework adapters. Native adapters are available for popular agent frameworks, and SDKs are provided for Python and TypeScript, with beta adapters for LangGraph and LangChain. This drop-in approach allows organizations to instrument their agents without refactoring existing systems.
Kaplaix offers a tiered pricing model: a Free plan with full event ingestion, risk scoring, session investigation, basic audit export (JSON), community support, and limited usage (one agent, one API key, 1,000 events per day); a Pro plan at $149 per month that adds PDF audit exports, team access and roles, webhook alerts, email support, and increased limits (up to 10 agents, 10 API keys, 50,000 events per day); and an Enterprise tier with custom governance options for regulated environments. No credit card is required to start using the Free plan.
In the LLM eval & observability space, kaplaix takes a focused approach. It focuses on providing tools for managing and governing AI agents in production environments. kaplaix is an open-source project aimed at developers building or deploying AI agents. The project is open source (MIT). It ships for the web, the command line, and API.
kaplaix first shipped in 2026. Among its 4 catalogued features are agent governance, observability tools, and Python SDK.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do