Noma Security

Noma Security makes an agent security and governance platform for enterprises adopting AI. It covers discovery, posture management, access control, adversarial testing, and runtime detection across endpoint agents, SaaS agent platforms, and homegrown AI applications.

What It Is

Noma is a unified AI security platform built around four products: AI-SPM, Agent Access Control, AI Red Teaming, and AI-DR. Security teams use it to find the agents, models, MCP servers, skills, and tools running across cloud, SaaS, and developer environments, then set rules for what those agents may do and watch their behavior in runtime.

How the Four Products Work Together

• AI-SPM maps each agent's connections, permissions, and data access (its blast radius) and surfaces misconfigurations and risky combinations of capabilities.
• Agent Access Control defines which MCP servers, skills, and tools agents may use, keyed to IdP groups or applied organization-wide, and blocks anything not approved.
• AI Red Teaming runs multi-turn attack campaigns against AI apps and agents for prompt injection, jailbreak, data leakage, and goal drift, with prebuilt scan profiles and compliance presets.
• AI-DR tracks the full chain of an agent session and detects prompt injection, data exfiltration, and scope violations. Depending on policy, it alerts, blocks, masks data, or routes to a human.

Noma says findings from one product feed the others, for example red teaming results becoming enforced runtime patterns.

Where It Fits

Noma targets three environments: employee-run endpoint agents such as coding assistants, SaaS agent platforms such as Copilot Studio and AgentForce, and engineering-built agents on platforms like Bedrock, Azure AI Foundry, and Databricks or frameworks like LangChain and CrewAI. Its Open Enforcement approach defines policy once and applies it at existing control points: agent hooks, MCP gateways, AI gateways, agent SDKs, and direct APIs. Solution pages cover agentic identity, MCP security, AI supply chain security, coding assistant security, and compliance reporting.

Vendor Claims

Noma states it often finds many times more agents than teams expect, and that its detection rates run 2-3x better than native cloud guardrails in POC benchmarks. These are vendor claims. The homepage also says Noma was named a Market Shaper in the 2026 Gartner Emerging Market Quadrant for AI Application Security.