NeuralTrust

NeuralTrust is an AI gateway that sits between your company's AI agents and the models, tools, and other agents they talk to. Instead of letting every employee or agent manage its own security and access, NeuralTrust gives you one central place to route calls, enforce policies, and log everything. It works with models from OpenAI, Anthropic, Azure OpenAI, and self-hosted options like vLLM — and you can switch between them without changing your agents' code.

The gateway is built around three core capabilities. The AI Gateway handles model routing, prompt inspection, PII redaction, and semantic caching — with under 100ms added latency at p95. The MCP Gateway lets you scope tool access per user, so an agent can read Salesforce but not write to the payments API. The A2A Gateway traces agent-to-agent handoffs, so when one agent delegates a task to another, every step is recorded. NeuralTrust also offers split-plane data residency, meaning sensitive data can stay inside your own perimeter while the control plane runs in the cloud.

NeuralTrust is open source and can be deployed as SaaS, hybrid cloud, or fully on-premises/air-gapped. It's built for security teams and AI engineering teams in regulated industries like finance, healthcare, and government. Use cases include centralizing audit trails for AI agents, enforcing least-privilege access to tools, and stress-testing chatbots before production go-live. The company recently raised $20M and counts ABANCA among its customers.