Protect AI

Protect AI is an AI/ML security platform (now part of Palo Alto Networks) whose products secure the AI lifecycle from model selection to runtime. Its developer surface centers on LLM Guard, an open-source Python toolkit of prompt and output scanners that ships a self-hostable REST API for real-time input/output sanitization. Commercial products - Guardian (model scanning), Recon (LLM red-teaming), and Layer (runtime protection) - are delivered through a portal under sales-led terms.

4 APIs 0 Features

AIMLSecurityLLMGuardrails

APIs

LLM Guard (Scanners)

Open-source (Apache 2.0) toolkit of input and output scanners that detect, redact, and sanitize LLM prompts and responses for prompt injection, PII, toxicity, secrets, and more....

Guardian (Model Scanning)

Commercial AI model security product that scans first- and third-party models for serialization attacks, malicious code, and supply-chain threats before they reach production. T...

Recon (Red-Teaming)

Commercial automated red-teaming product that rigorously tests LLM and GenAI applications for vulnerabilities, jailbreaks, and policy violations. Delivered as a sales-led platfo...

Layer (Runtime)

Commercial runtime security product that monitors and controls AI applications in production with deep visibility and inline threat prevention. Delivered as a sales-led platform...

Collections

LLM Guard API

OPEN

Pricing Plans

Protectai Plans Pricing

3 plans

PLANS

Rate Limits

Protectai Rate Limits

2 limits

RATE LIMITS

FinOps

Protectai Finops

FINOPS

Resources

👥

GitHubOrganization

Sources

opencollection: 1.0.0
info:
  name: LLM Guard API
  version: '1.0'
request:
  auth:
    type: bearer
    token: '{{bearerToken}}'
items:
- info:
    name: System
    type: folder
  items:
  - info:
      name: Service metadata.
      type: http
    http:
      method: GET
      url: http://localhost:8000/
    docs: Returns basic service and version information for the running LLM Guard API.
  - info:
      name: Liveness probe.
      type: http
    http:
      method: GET
      url: http://localhost:8000/healthz
    docs: Returns the liveness status of the API process.
  - info:
      name: Readiness probe.
      type: http
    http:
      method: GET
      url: http://localhost:8000/readyz
    docs: Returns readiness once scanners are loaded and the API can serve scan requests.
  - info:
      name: Prometheus metrics.
      type: http
    http:
      method: GET
      url: http://localhost:8000/metrics
    docs: Exposes Prometheus-format metrics for the API when metrics are enabled.
- info:
    name: Prompt
    type: folder
  items:
  - info:
      name: Analyze and sanitize a prompt.
      type: http
    http:
      method: POST
      url: http://localhost:8000/analyze/prompt
      body:
        type: json
        data: "{\n  \"prompt\": \"<string>\",\n  \"scanners_suppress\": []\n}"
    docs: Runs the configured input scanners over the prompt and returns sanitized_prompt, is_valid, and per-scanner risk
      scores.
  - info:
      name: Scan a prompt.
      type: http
    http:
      method: POST
      url: http://localhost:8000/scan/prompt
      body:
        type: json
        data: "{\n  \"prompt\": \"<string>\",\n  \"scanners_suppress\": []\n}"
    docs: Runs the configured input scanners over the prompt and returns is_valid and per-scanner risk scores without modifying
      the prompt.
- info:
    name: Output
    type: folder
  items:
  - info:
      name: Analyze and sanitize an output.
      type: http
    http:
      method: POST
      url: http://localhost:8000/analyze/output
      body:
        type: json
        data: "{\n  \"prompt\": \"<string>\",\n  \"output\": \"<string>\",\n  \"scanners_suppress\": []\n}"
    docs: Runs the configured output scanners over the LLM response and returns sanitized_output, is_valid, and per-scanner
      risk scores.
  - info:
      name: Scan an output.
      type: http
    http:
      method: POST
      url: http://localhost:8000/scan/output
      body:
        type: json
        data: "{\n  \"prompt\": \"<string>\",\n  \"output\": \"<string>\",\n  \"scanners_suppress\": []\n}"
    docs: Runs the configured output scanners over the LLM response and returns is_valid and per-scanner risk scores without
      modifying the output.