🚨 NeuralTrust has raised $20M

The Agent Gateway built for security

Connect your company AI agents to the models and tools they need — and centralize security, observability, and governance over every call.
TrustGate AI gateway dashboard showing activity logs, request inspection and the plugin chain
Trusted by the world’s leading companies
Architecture

Security and governance on every call

Without a gateway, every employee secures and controls its own AI traffic — or no one does. NeuralTrust turns thousands of individual decisions into one you control.
TrustGate as a single control point between your apps and agents and the models, tools and agents they use
Capabilities

From one agent to thousands

Enable agents across your organisation — with central enforcement over what every agent can reach and a record of every call.

Reach any model, tool, or agent

AI Gateway

Use any model, switch anytime

Point your agents at any model — OpenAI, Anthropic, or self-hosted. Every call routes through one place you control and meter.

Model routingActive
AIOpenAI · gpt-4o
AnAnthropic · claude
AzAzure OpenAI
⌂Self-hosted · vLLM
MCP Gateway

Give employees secure access to thousands of tools

Let your agents act across the tools your business runs on — with access scoped per user and every call on record.

Tool access · per user12 scoped
â–¦Salesforceread
▤GitHubread · write
â—«Payments APIdenied
â–¥Internal wikiread
A2A Gateway

Let agents work together, under your watch

Your agents can delegate to each other as workflows grow. Every handoff is traced, nothing happens off the record.

Handoff tracelive
A1Planner
A2Worker
A3Reviewer
Delegated: summarise contract142ms
Tool call: doc.search()88ms
Approved & traced12ms
Performance

Enterprise performance with security in the data path

Battle-tested inside the most regulated environments. Inline prompt inspection, PII redaction, and semantic caching — with low overhead for production traffic.
Latency
<100ms
Added at p95with PromptGuard and Data Masking enabled
Throughput
20k+
Requests / second / nodetested under sustained enterprise load
Detection
99%
Multilingual accuracyjailbreaks, injections, policy violations
Sovereignty
100%
Split-plane data residencysensitive data stays in your perimeter
Use cases

Why an Agent Gateway?

Without a Gateway, agentic workloads leave security gaps, observability blind spots, and no audit trail your team can defend.

Identity
The problem

Agents act on behalf of users

They invoke tools dynamically, and chain across services. API-level auth was never designed to track who actually authorized what.

With TrustGate

Identity as a first-class citizen

End-user identity is forwarded through every hop. Per-agent and per-tool RBAC. Cryptographic audit trails that survive review by your security team and your auditor.

Security overviewMCP authorizationZero-trust for agents
Open source

Our AI Gateway is open source. Give it a try

Run TrustGate locally or self-hosted in your own cloud in minutes. Start from the open core, then upgrade to fit your organisation's governance, compliance, and depth-of-control needs across many teams.

$curl -fsSL https://neuraltrust.ai/install.sh | bash
Deployment

Deploy where compliance requires

TrustGate runs wherever your data lives. No architecture changes, no security exceptions.

SaaS

Fastest to start. Managed infra, automatic upgrades.

Up in under an hour
Shared tenancy (data isolated)
ISO27001 certified

Hybrid cloud

SaaS control plane, privately hosted data plane. Data never leaves your boundary.

Single-tenant
HIPAA & GDPR ready
Custom data residency

On-premises / air-gapped

Full isolation. For defence, government, and financial sector.

No external calls
FedRAMP & IL4 pathway
Kubernetes + Helm chart
Customers

Trusted by security leaders

Juan Manuel Sanchez-Quinza

With NeuralTrust we stress-tested our chatbot with GenAI ‘SOFia,’ validating a safe go-live that meets financial-sector security and regulatory standards.

Director of Transformation, ABANCA

See the Agent Gateway in action

Get a demo
Ai Gateway | NeuralTrust