Documentation — Kovral.ai

Quick Start Proxy Integration SDK Install Writing Policies Agent Security Webhooks API Reference

Quick Start

The fastest integration is the proxy method — change one line of code and every AI call you make is immediately protected.

# Before
client = OpenAI(base_url="https://api.openai.com/v1")

# After — full protection, zero other changes
client = OpenAI(
  base_url="https://api.kovral.ai/v1",
  default_headers={"X-Kovral-Key": os.environ["KOVRAL_API_KEY"]}
)

Writing Policies

Policies are defined in a simple YAML format using plain-language conditions. No rules-engine expertise required.

policies:
  - name: block-pii-external
    description: "Block PII in prompts to third-party models"
    when:
      content_contains: [pii, ssn, credit_card]
      destination: external
    action: block
    alert: slack://security-alerts

  - name: coach-shadow-ai
    description: "Alert on unsanctioned tools"
    when:
      tool_status: unsanctioned
    action: allow
    notify_user: true
    message: "This tool hasn't been reviewed by security yet."

Webhooks

Kovral sends a POST to your endpoint for every threat event. Payload example:

{
  "event": "threat.detected",
  "severity": "critical",
  "threat_type": "prompt_injection",
  "user_id": "usr_7x3k9p",
  "tool": "ChatGPT",
  "policy_matched": "block-injection-attempts",
  "action_taken": "blocked",
  "timestamp": "2024-10-09T14:32:07Z",
  "request_id": "req_ab3kx92mn"
}

Social Proof

Trusted by Security Teams Worldwide

From Fortune 500 CISOs to fast-growing AI startups — teams protecting AI at scale choose Kovral.

★★★★★

"Kovral gave us our first real-time view into every AI tool running inside our bank. We found 34 unsanctioned models in the first 48 hours. The posture score is now a standing item on my CISO dashboard."

Marcus Reid

VP Security Engineering · Tier-1 Bank

★★★★★

"Our EU AI Act readiness went from 'we don't know' to a documented compliance posture in three weeks. The use case registry and automated evidence collection saved our team months of manual work."

Priya Nair

Chief Compliance Officer · HealthTech SaaS

★★★★★

"The AI Firewall blocked a live indirect prompt injection attack on our customer-facing agent within the first week of deployment. Zero code changes, 1.8ms added latency. It's the easiest security win I've had in years."