Fabraix offers Arx
Product
Arx is Fabraix's run-time defence layer that blocks prompt injection and memory poisoning in AI agents.
- Currently Offered
Arx is a run-time defence layer for AI agents, built from the offensive findings of Fabraix's Nyx agent. It registers agent sessions under a trace_id, records user inputs, model outputs, tool calls, memory operations, and environment events, and validates proposed actions against the full session context before they execute.
Arx is consumed directly through the api.fabraix.com/v1 REST API with an x-api-key header; no SDK is required, and a Python SDK is listed as roadmap. Its documented endpoints are POST /register-agent-run, POST /event, and POST /check.
Arx is Fabraix's run-time defence layer that blocks prompt injection and memory poisoning in AI agents.
Open-source Python framework that runs configurable Input/Output Guards to validate large language model responses.
Stateful API sandboxes for testing AI agents without hitting live third party services.
Amazon Bedrock Guardrails is a configurable safeguard layer that evaluates and filters inputs and outputs of generative AI applications.
Open-weight security triage model scoring agent, application and infrastructure events with calibrated probabilities.
Identity-aware network access layer that replaces long-lived AI API keys for LLM providers and AI agents.
AI Control Plane monitors, evaluates, and governs enterprise AI agents across their lifecycle.
AI Agent Reliability Platform detects and remediates incorrect AI agent responses in real time.
Enterprise platform for building, deploying, observing, and governing production AI agents.
Aedeon Agent Fabric stands up specialized governed agents for modernization tasks.
Hosted gateway for routing and governing traffic across hosted and self-hosted AI models.
Enterprise AI observability and evaluation platform for testing, monitoring, and improving AI agents and applications.
AI copilot for debugging, optimizing, and analyzing LLM and agent performance.
Lakera Guard is a runtime security product for large language model applications.
Open-source Python framework that runs configurable Input/Output Guards to validate large language model responses.
Stateful API sandboxes for testing AI agents without hitting live third party services.
Amazon Bedrock Guardrails is a configurable safeguard layer that evaluates and filters inputs and outputs of generative AI applications.
Open-weight security triage model scoring agent, application and infrastructure events with calibrated probabilities.
AgentOps is an observability platform for tracing, debugging, replaying, and monitoring AI agent activity.
Identity-aware network access layer that replaces long-lived AI API keys for LLM providers and AI agents.
AI Control Plane monitors, evaluates, and governs enterprise AI agents across their lifecycle.
AI Agent Reliability Platform detects and remediates incorrect AI agent responses in real time.
Fugu is a multi-agent orchestration API that routes complex tasks across a pool of frontier language models.
Enterprise platform for building, deploying, observing, and governing production AI agents.
Aedeon Agent Fabric stands up specialized governed agents for modernization tasks.
Vanta Agent is the 24/7 AI GRC engineer for policies, questionnaires, and TPRM.
memgram provides AI-agent memory infrastructure for development teams, with inspectable memory state and traceable retrieval.
Hosted gateway for routing and governing traffic across hosted and self-hosted AI models.