Enkrypt AI Guardrails is a runtime control layer for AI agents, tools, retrieval-augmented generation and MCP. It can filter, rewrite, block or escalate behavior at prompt, retrieval, tool and output boundaries, including approving or denying tool calls. The product addresses risks such as prompt injection, unsafe tool actions, privilege or tenant boundary violations, sensitive data exposure, jailbreaks, toxic content and compliance issues. Policies apply to text, image and audio inputs. Integration options include existing model stacks, agent hooks or middleware, an MCP Gateway, SIEM and ticketing routes, and identity claims from Okta, Azure AD or custom JWT claims. Enforcement logs include a policy ID, version and reason code and can be exported to SIEM. The product page claims latency under 15 milliseconds. The free Explore plan includes 50 credits per month and seven-day data retention; Launch is $134 per month billed annually, with 30-day retention. Enterprise offers VPC or on-premises deployment and custom retention.
Who it is for
It suits enterprises securing AI deployments, as well as AI safety researchers and people building safer AI. Explore is described as suitable for spot checks and evaluation, while comprehensive red-team assessment requires 5,000 credits.
What is good
- Controls apply at prompt, retrieval, tool and output boundaries.
- Policies cover text, image and audio inputs.
- Logs include policy IDs, versions and reason codes.
- Free Explore plan includes 50 credits monthly.
- Enterprise offers VPC or on-premises deployment.
What to know first
- Explore data retention is seven days.
- Explore includes 50 credits per month.
- Comprehensive red-team assessment requires 5,000 credits.
- Enterprise pricing requires contacting sales.
Verdict
Enkrypt AI Guardrails offers multiple enforcement points and records decisions for audit evidence. Compare the Explore credit and retention limits with your evaluation needs; paid Launch pricing is $134 per month billed annually.
Enkrypt AI Guardrails plans and pricing
All plansCompared on AI guardrail software
- Free plan
- Yes
- Paid from
- $134/mo
- Prompt injection defense
- Yes
- PII detection
- Yes
- Jailbreak detection
- Yes
- Custom policies
- Yes
- SDK languages
- Python
