SPLX

SPLX is an AI security platform, now part of Zscaler according to the site banner, that helps enterprises test, protect and govern AI systems from development through deployment. It combines automated red teaming, runtime guardrails, asset discovery, compliance mapping and remediation. SPLX also publishes Agentic Radar, an open-source security scanner for agentic workflows.

What It Is

SPLX is a unified platform for securing the AI lifecycle. Security teams connect AI systems through connectors (REST API, conversational platforms and large language models, with no coding required, per the site) and then run testing, monitoring and governance workflows against them. The company describes the platform as built by AI red teamers for enterprise scale.

Platform Capabilities

• AI Asset Management: discovers models, AI workflows, MCP servers and guardrails and builds an AI-BOM.
• Automated AI Red Teaming: continuously tests AI systems using an attack database and custom datasets.
• AI Runtime Protection: applies input/output guardrails to detect prompt injection, jailbreaks and sensitive data leakage.
• AI Runtime Threat Inspection: scans LLM logs to detect malicious or unsafe queries.
• AI Governance & Compliance: maps findings to global and custom security frameworks with audit-ready reporting.
• Dynamic Remediation: generates hardened system prompts and remediation steps from red teaming results.
• AI Model Security: benchmarks commercial and open-source LLMs with adversarial prompts to inform model selection.

Agentic Radar (Open Source)

Agentic Radar is an Apache-2.0 Python CLI installed with pip. Its scan command analyzes code for agentic workflows (LangGraph, CrewAI, n8n, OpenAI Agents, Autogen) and generates an HTML report with a workflow graph, tool identification, MCP server detection and vulnerability mapping to the OWASP Top 10 for LLM Applications and OWASP agentic threat guidance. Optional prompt hardening rewrites detected system prompts, and a test command runs adversarial tests such as prompt injection and PII leakage against OpenAI Agents workflows, with custom YAML tests supported. A GitHub Actions example enables CI/CD scanning. Static analysis runs locally; optional features may send prompts to an LLM.