Enkrypt AI Guardrails

Enkrypt AI Guardrails is a runtime layer for enforcing policies across AI agents and the tools, retrieval systems, and models they use. It is best suited to teams securing an existing AI deployment that need controls at several points in a workflow. Its broad enforcement options are a clear strength, but credit allowances and retention on lower plans constrain evaluation and ongoing review.

Overview

The product can approve, change, or block risky behavior at prompt, retrieval, tool, and output boundaries. That wider coverage is useful for agent systems, where risk can arise from a tool action or retrieved data as well as from the final response. Controls address prompt injection, jailbreaks, unsafe tool actions, privilege or tenant boundary violations, sensitive data exfiltration, toxic content, and compliance risks.

Policies also cover image and audio inputs, including injection defenses. That makes Enkrypt relevant to multimodal workflows, not only text-based applications. It is available through API, self-hosted, and web platforms.

Key features

Workflow enforcement and integrations

Filtering, rewriting, blocking, escalation, and tool-call approval or denial give teams several ways to respond to policy violations. API-first integration, agent hooks or middleware, and an MCP Gateway offer routes into existing model and tool stacks. Identity claims can use Okta, Azure AD, or custom JWT claims, which helps teams account for identity context when applying controls. Python is the listed SDK language.

Audit, performance, and security

Enforcement logs record a policy ID, version, and reason code, and can be exported to SIEM. Those records can support internal control evidence and review, though their usefulness over time depends on the retention limit of the selected plan. The product page claims latency under 15 milliseconds and stable behavior under load; these are vendor claims rather than a guarantee for every deployment.

The pricing page lists SOC 2 Type II, ISO 27001, GDPR Ready, HIPAA Ready, and NIST AI RMF Aligned. Support ranges from community and email options to dedicated Slack or Teams and Enterprise 24x7 support availability.

Pricing

Explore is free, billed Free forever. It starts with 500 credits, adds 50 credits per month, includes 7-day data retention, and has community support. That is enough for spot checks and evaluation, but not a broad red-team assessment: a comprehensive assessment requires 5,000 credits.

Launch is $134 a year, billed annually. It provides 5,000 credits to start, 250 credits per month, 30-day retention, and email support. It suits modest workloads, but the monthly allocation remains limited for sustained, extensive testing. Scale is $1,349 a year, billed annually, with 10,000 starting credits, 1,000 credits per month, 30-day retention, and dedicated Slack or Teams support. It is the stronger fit for teams with higher recurring usage that do not need a custom deployment.

Enterprise has custom pricing, custom allocation and retention, custom SLAs, and unlimited usage for VPC. It also includes VPC or on-premises deployment, making it the plan for organizations with deployment or service requirements the lower tiers do not cover. No seat allowances are stated for these plans.

Who it's for

Enkrypt is aimed at enterprises securing AI deployments, AI safety researchers, and people building safer AI. It is the better fit when a team needs controls across prompts, retrieval, tools, and outputs, plus exportable decision records. A team that only needs a small monthly volume of checks may find the free plan adequate; teams planning comprehensive evaluation should account for the 5,000-credit requirement.

Pros and cons

• Pro: Enforcement reaches prompts, retrieval, tool calls, and outputs, addressing risks beyond the final generated answer.
• Pro: Image and audio policy coverage extends protection to multimodal inputs.
• Pro: Exportable logs with policy identifiers and reason codes support audit and internal review.
• Con: Explore's 50 monthly credits and 7-day retention make it a narrow evaluation tier rather than a durable operational plan.
• Con: Even Launch's 250 monthly credits may be restrictive for teams running frequent checks or extensive assessments.
• Con: VPC or on-premises deployment and unlimited VPC usage require Enterprise, with custom pricing.

Alternatives

AI Guardrail Software is the category list for comparing tools across this space.

• PromptGuard is worth considering when a large free scan allowance matters more than Enkrypt's broader workflow coverage: its Free plan includes 20,000 scans a month, one API key, one project, and 24-hour log retention.
• OpenSecureAI Prompt Firewall may suit readers who want a free self-hosted or web option with in-browser tools and npm packages; its Free plan includes 2,000 API requests per month and a BYOK LLM Gateway.
• Fiddler Guardrails is another freemium API, self-hosted, and web option, with a free plan covering real-time detection of harmful exposure, hallucinations, toxicity, PII/PHI, prompt injection, and jailbreak attempts.
• Amazon Bedrock Guardrails is a paid web option for readers considering a product with separately priced content filters and denied topics.
• Openlayer Guardrails has a free Basic plan with one member, five projects, one inference pipeline per project, and 20,000 inferences per month, making it an alternative to consider for those plan limits.
• HiddenLayer AI Runtime Security is a paid alternative available across API, Linux, self-hosted, and web platforms.
• OpenAI Guardrails is a free API, self-hosted, and web alternative.
• Pangea AI Guard is a freemium option with a free plan across API, extension, and self-hosted platforms.

Verdict

Choose Enkrypt AI Guardrails if your team needs runtime policy enforcement across agent workflows and auditable decisions, especially when multimodal inputs or tool actions are part of the risk surface. Its breadth is the main reason to choose it; limited credits and retention on lower tiers, and custom pricing for enterprise deployment, are the main reasons to look elsewhere.