Products

One platform, three ways to run it

AsrielNetworks AI XDR is the product. The gateway, the SDK and the evaluation lab are how it reaches your systems. Every tier runs the same detection engine on the same data model; tiers differ in what the console exposes and how far correlation and response go.

Compare tiers Sign in

The AI Incidents queue: incidents ranked by risk score with severity, status, containment, recovery steps, agent and model, attack class and analyst. — AI Incidents, the workspace every finding lands in.

Primary product

AsrielNetworks AI XDR

Extended detection and response for agents, LLM applications and MCP hosts: telemetry ingest, deterministic detection over prompt-event traces, incident correlation and response actions, in one tenant-isolated console.

What it does

Ingests asrielnetworks.ai.event.v1 telemetry from the SDK, a runtime path, or the inline gateway.
Runs the twelve built-in rules over each trace: prompt injection, excessive agency, data access abuse, supply chain, causal hijack.
Correlates findings into incidents with risk scores, attack paths, evidence and response actions.
Keeps every tenant's rows behind forced row-level security and an append-only audit log.

Who it is for

Security engineering teams shipping agents that call tools.
SOC analysts who need incidents, not model logs.
Platform teams that own the model gateway and want detection in the request path.
Compliance owners who need governance evidence for AI systems.

Get it

Three questions about your deployment (where your agents run, what you will use it for, which tier), then your console. No account yet? You create one on the way; your answers are kept for it.

How to use it, module by module: Free · Pro · Enterprise

Tiers

What each tier exposes

Gates are enforced on the API, not only in the console: a route a tier does not include answers 403 plan_limit. Pro and Enterprise start as 14-day trials.

Console features by tier
Capability	Free	Pro	Enterprise
Monitoring, alerts, notifications; all built-in detection rules	included	included	included
Attack stories and incident attack path	—	included	included
AI co-pilot over your own telemetry	—	included	included
AI firewall rules and case management	—	included	included
Advanced analytics: attack map, users map	—	included	included
AI threat intelligence, governance dashboard, tool & MCP visibility	—	included	included
Playbooks and custom detection rules	—	—	included
Causal capability firewall: causal trace and counterfactual replay	—	—	included
Secure terminal, team management, SSO (SAML / OIDC)	—	—	included
Organization exports and evaluation lab	—	—	included

Or talk to sales

Each plan has a guide to its modules: Free · Pro · Enterprise.

Inside the console

The screens an analyst lives in

Captured from the console on its demonstration dataset. Every page reads the same API your scripts can.

AI Agents: each agent with status, tasks today, blocked actions, the model it runs on and a risk score out of 100. — AI Agents: memory, permissions, tools, autonomy and a behaviour history per agent.

Tools and MCP: connected MCP servers with status, permission, last call, calls in 24 hours, bound agents, and a blocked server with the unauthorized command the broker refused. — Tools & MCP: every server, every permission, every call the broker allowed or refused.

Prompt and Event Explorer: prompts, retrieved chunks, tool calls, MCP events and outputs in one searchable index. — Prompt & Event Explorer: one query across every AI event.

Detection coverage: MITRE ATLAS and ATT&CK techniques with coverage status, the detection that covers each, and activity in 24 hours. — Detection coverage: MITRE ATLAS and ATT&CK techniques against the rules that cover them.

Also in the platform

How the platform reaches your systems

These ship with AI XDR; each is a different place to attach detection. New products are listed here as they are released.

Inline model gateway

An OpenAI-compatible proxy at /gateway/v1 for chat completions, the Responses API and embeddings. Uses your provider credential, records every request as prompt events, detects on the way through, and pins egress to the endpoint validated when the deployment was registered.

For: platform teams that already own the base URL their applications call.

SDK and runtime telemetry

The asrielnetworks.ai.event.v1 wire contract, a batch ingest route, and a runtime route for single events. Machine credentials are scoped to an application, agent and deployment; a rejected event does not sink its batch.

For: teams instrumenting their own agent loop, or a host that cannot route through a proxy.

See the event shape

Evaluation lab

Active probes that drive a registered deployment through the gateway with adversarial cases and report which rules fired, so a change to a prompt, a model or a tool set is measured before it ships.

For: teams that gate releases on security behaviour. Enterprise tier.

Talk to sales