Enkrypt AI Agent Red Teaming and Guardrails

Enkrypt AI, Inc. is an AI security and compliance company based in Brighton, Massachusetts, led by co-founder and CEO Sahil Agarwal. It helps enterprises secure, monitor and govern their AI agent stack, publishes an LLM safety leaderboard ranking more than 200 models, and was named a Gartner Cool Vendor in AI Security in 2025. In August 2026 Anaconda announced it is acquiring the company.

The platform covers the agent lifecycle. Agent Red Teaming runs adversarial tests on models and agents across text, image and voice; Agent Guardrails protect production systems at runtime; and the Agent Policy Engine turns written policies into enforceable guardrails. The AI Data Risk Audit checks training data, fine-tunes and embeddings, while the MCP Gateway and MCP Scanner control and assess MCP servers. Compliance features map controls to the EU AI Act and NIST and produce evidence, and an AI asset scanner and third-party AI risk module support governance. Case studies include AI21 Labs and NATO StratCom COE.

• Agent Red Teaming: adversarial testing of models and agents across text, image and voice.
• Agent Guardrails: runtime protection against unsafe inputs and outputs.
• Agent Policy Engine: converts written policies into enforceable guardrails.
• AI Data Risk Audit: checks training data, fine-tunes and embeddings.
• MCP Gateway and Scanner: one controlled endpoint and risk scanning for MCP servers.
• Compliance and audit: control mapping to the EU AI Act and NIST with evidence generation.

Vietnamese AI startups can use the free or Launch tier to red-team chatbots before pitching to bank or telecom customers. Banks can apply guardrails and policy enforcement to customer-facing agents. Companies connecting agents to ERP or CRM through MCP can scan servers for risky tools.

Teams can start with a free tier and scale to paid plans, making AI red teaming affordable for smaller companies. Automation shortens the path to certify and ship AI systems, which the company says can be cut by 90%. Policies written in plain language become runtime guardrails, closing the gap between compliance teams and engineers. MCP scanning and gateway reduce risk from third-party agent tools.

Enterprises are moving from chatbots to agents that call tools and MCP servers, multiplying the attack surface. Open-source AI platforms are also bundling security, as Anaconda's acquisition shows. Vietnamese startups and enterprises building agents on open models need affordable red teaming and guardrails that scale with usage.

Based in Brighton, Massachusetts, USA. Raised a 2.35 million USD seed round in February 2024. Named a Gartner Cool Vendor in AI Security in 2025, won a 2025 Cyber Security Excellence Award and was an Accenture Ventures 2025 winner in AI infrastructure. In August 2026 Anaconda announced it is acquiring Enkrypt AI; the Enkrypt AI site and products remain online.