NeuralTrust is an AI security company that protects enterprises as they put AI agents into production. Its platform sits in the path of every call an agent makes, to models, MCP servers, tools and other agents, and inspects prompts, responses and tool calls as they happen.
It is aimed at security and platform teams that need to know which agents are running, control what each one may reach, and stop attacks such as prompt injection before they cause harm.
At A Glance
| Vendor | NeuralTrust |
|---|---|
| Category | AI agent runtime security and AI gateway |
| Identities covered | End users, AI agents, MCP tools |
| Deployment | SaaS, hybrid cloud, or on-premises and air-gapped |
| Agentic AI | MCP Gateway, A2A Gateway, per-agent and per-tool RBAC, runtime detection |
| Certifications | ISO 27001 (SaaS) |
| Customers | Sabadell, Iberia, ABANCA, ISDIN, Air Europa |
| Market classification | AI Safety & Guardrails; also AI Governance & Compliance, Agentic AI Identity |
What NeuralTrust Does
When each team wires its own agents to models and tools, AI traffic ends up governed inconsistently or not at all. NeuralTrust’s answer pairs a gateway that every agent call passes through with a runtime layer that inspects each interaction. The work runs in four stages, discover, evaluate, respond and enforce, so agents are found and tested before release and then governed at runtime with events sent to the customer’s SIEM. One policy then covers every team’s agents, whichever model or tool they use.
Products
Agent Runtime Security (TrustGuard)
TrustGuard is for teams that need to stop attacks at the moment an agent acts. It inspects every interaction against detection rules for injections, sensitive data, unsafe actions and behavioural attacks, then allows, blocks or transforms the traffic, or simply raises an alert. Detection is stateful, so it follows a conversation across turns and catches probing or slow escalation that a single message would not reveal.
Agent Gateway (TrustGate)
TrustGate gives an organisation one control point for agent traffic.
| AI Gateway | Routes model calls to OpenAI, Anthropic, Azure OpenAI or self-hosted vLLM, with AI cost management and semantic caching. |
|---|---|
| MCP Gateway | Lets agents use business tools with access scoped to each user and every call on record. |
| A2A Gateway | Traces each handoff when one agent delegates to another. |
| Data protection | PromptGuard, data masking and PII redaction on traffic in flight. |
AI Red Teaming (TrustTest)
TrustTest is a red-teaming and evaluation toolkit for teams that want to find weaknesses before attackers do. It targets an HTTP endpoint, an LLM provider or a custom target and covers jailbreaks, prompt injection including indirect injection, system-prompt extraction, data leaks, hallucination, bias and agent and tool abuse. Results are graded with heuristics and LLM-as-judge. It runs from the terminal, as a CI/CD gate that blocks a merge, or as scheduled continuous runs, drawing on an attack catalogue of more than 150 entries.
Agent Posture Management
Posture management gives a view of the agents, models and tools in use and where they are exposed, with more than 1.7 million models reported scanned.
How It Works
The platform separates a control plane from a data plane and offers three deployment models.
| Component | What it does |
|---|---|
| SaaS | Managed infrastructure with automatic upgrades and shared tenancy with isolated data, certified to ISO 27001. |
| Hybrid cloud | A SaaS control plane with a privately hosted, single-tenant data plane, so data stays inside the customer’s boundary. It is ready for HIPAA and GDPR. |
| On-premises and air-gapped | Full isolation for defence, government and financial sector customers, deployed on Kubernetes with a Helm chart and a FedRAMP and IL4 pathway. |
Collectors take in agent traffic from gateways, SDKs, browsers, sidecars and log streams. NeuralTrust does not train on customer data.
Identities And Systems Covered
| End users | The end user’s identity is carried through every hop of gateway traffic. |
|---|---|
| AI agents | A live inventory of deployed agents, with per-agent RBAC. |
| Tools and MCP servers | Per-tool RBAC and MCP access scoped to each user. |
| Models | OpenAI, Anthropic, Azure OpenAI and self-hosted vLLM. |
| AI tools in use | ChatGPT, Gemini, Microsoft Copilot, Cursor and Claude Code. |
Agentic AI Capabilities
Identity-Aware Decisions
Every runtime decision weighs the identity behind the request, the protocol in use and the collector metadata, so the same prompt can be treated differently for different users or agents. Because the end user’s identity travels with each hop, an agent acting for a user cannot reach a tool that user could not reach, even when one agent delegates to another.
Audit
The gateway keeps cryptographic audit trails of agent activity for security and audit review. Across deployments, the company reports more than 3 million attacks blocked.
Company Background
| Founded | 2022, in Barcelona. |
|---|---|
| Leadership | Joan Vendrell (CEO), Victor García (CTO) and Alejandro Domingo (COO), the founders. |
| Offices | Barcelona, New York and London. |
| Funding | A reported $20 million round led by Alstin Capital, with Seaya and Kibo taking part; other investors include VentureFriends, Banco Sabadell, Plug and Play and IESE. |
| Recognition | Included in the 2026 Gartner Hype Cycle for Secure AI as a sample vendor for AI Runtime Defense. |
Note: This overview is compiled from NeuralTrust’s publicly available information as of 4 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.