Lakera

aid: lakera-ai
url: https://raw.githubusercontent.com/api-evangelist/lakera-ai/refs/heads/main/apis.yml
apis:
  - aid: lakera-ai:lakera-guard-api
    name: Lakera Guard API
    tags:
      - AI Security
      - Artificial Intelligence
      - Guard
      - Prompt Injection
      - PII
      - Content Moderation
    humanURL: https://docs.lakera.ai/docs/api/guard
    properties:
      - url: https://docs.lakera.ai/docs/api/guard
        type: Documentation
      - url: https://docs.lakera.ai/docs/quickstart
        type: GettingStarted
      - url: https://docs.lakera.ai/api-reference/lakera-api/guard/screen-content
        type: Documentation
      - url: openapi/lakera-guard-api-openapi.yml
        type: OpenAPI
      - url: json-schema/lakera-guard-request-schema.json
        type: JSONSchema
      - url: json-schema/lakera-guard-response-schema.json
        type: JSONSchema
      - url: json-structure/lakera-guard-request-structure.json
        type: JSONStructure
      - url: json-ld/lakera-ai-context.jsonld
        type: JSONLD
      - url: examples/lakera-guard-screen-content-example.json
        type: Example
      - url: examples/lakera-guard-results-example.json
        type: Example
    description: >-
      Lakera Guard's runtime screening API. POST OpenAI-style chat messages to /v2/guard and receive a `flagged`
      decision based on the policy assigned to the project, with optional per-detector breakdown and payload-level
      PII/profanity match locations. Detectors cover prompt attacks, data leakage, PII, content violations, and unknown
      links across 100+ languages with sub-50ms latency. A second endpoint, /v2/guard/results, returns detector
      confidence levels (L1–L5) for offline analysis and threshold tuning without contributing to runtime screening
      logs.
  - aid: lakera-ai:lakera-platform-api
    name: Lakera Platform API
    tags:
      - AI Security
      - Administration
      - Policies
      - Projects
      - Enterprise
    humanURL: https://platform.lakera.ai/docs/api
    properties:
      - url: https://platform.lakera.ai/docs/api
        type: Documentation
      - url: openapi/lakera-platform-api-openapi.yml
        type: OpenAPI
    description: >-
      The Lakera Platform API lets Enterprise SaaS customers programmatically manage Lakera Guard policies and projects.
      Policies select detectors and per-detector sensitivity (L1–L5); projects bind a protected application to a policy
      and an API key scope. Self-hosted deployments expose additional /policies/lint and /policies/health endpoints plus
      Kubernetes startupz/readyz/livez probes for in-cluster operation.
name: Lakera
tags:
  - AI Security
  - Artificial Intelligence
  - Generative AI
  - LLM Security
  - Prompt Injection
  - AI Guardrails
  - AI Red Teaming
  - Data Loss Prevention
  - Content Moderation
  - Check Point
kind: contract
image: https://kinlane-images.s3.amazonaws.com/shared/apis-json/apis-json-logo.jpg
access: 3rd-Party
common:
  - type: LinkedIn
    url: https://www.linkedin.com/company/lakeraai
  - url: https://www.lakera.ai
    type: Portal
  - url: https://docs.lakera.ai
    type: Documentation
  - url: https://docs.lakera.ai/docs/api
    name: API Overview
    type: Documentation
  - url: https://docs.lakera.ai/docs/quickstart
    type: GettingStarted
  - url: https://docs.lakera.ai/guard
    name: Introduction to Lakera Guard
    type: Documentation
  - url: https://docs.lakera.ai/docs/integration
    name: Lakera Guard Integration Guide
    type: Documentation
  - url: https://docs.lakera.ai/llms.txt
    name: llms.txt index
    type: Documentation
  - url: https://docs.lakera.ai/_mcp/server
    name: Lakera Docs MCP Server
    type: MCPServer
  - url: https://platform.lakera.ai
    name: Lakera Platform
    type: Portal
  - url: https://platform.lakera.ai/pricing
    name: Pricing
    type: Pricing
  - url: https://www.lakera.ai/contact
    type: Support
  - url: https://www.lakera.ai/blog
    type: Blog
  - url: https://www.lakera.ai/news
    type: Press
  - url: https://www.lakera.ai/customers
    name: Customers
    type: CaseStudy
  - url: https://www.lakera.ai/legal/terms
    type: TermsOfService
  - url: https://www.lakera.ai/legal/privacy
    type: PrivacyPolicy
  - url: https://trust.lakera.ai
    type: TrustCenter
  - url: https://gandalf.lakera.ai
    name: Gandalf (prompt injection game)
    type: Sandbox
  - url: https://github.com/lakeraai
    type: GitHubOrganization
  - url: https://github.com/lakeraai/pint-benchmark
    name: PINT Benchmark
    type: Tool
  - url: https://github.com/lakeraai/guard-demo-client
    name: Lakera Guard Demo Client (Python)
    type: CodeExamples
  - url: https://github.com/lakeraai/chrome-extension
    name: Lakera Chrome Extension (ChatGPT Data Leak Protection)
    type: Tool
  - url: https://github.com/lakeraai/canica
    name: Canica Text Embedding Viewer
    type: Tool
  - url: https://github.com/lakeraai/dsec-gandalf
    name: dsec-gandalf
    type: Tool
  - url: https://github.com/lakeraai/intent-augmentation
    name: Intent Augmentation
    type: Tool
  - url: https://docs.litellm.ai/docs/proxy/guardrails/lakera_ai
    name: LiteLLM Lakera Guardrail
    type: Integration
  - url: plans/lakera-ai-plans-pricing.yml
    type: Plans
  - url: rate-limits/lakera-ai-rate-limits.yml
    type: RateLimits
  - url: finops/lakera-ai-finops.yml
    type: FinOps
  - url: rules/lakera-rules.yml
    type: SpectralRules
  - url: vocabulary/lakera-ai-vocabulary.yml
    type: Vocabulary
  - type: Features
    data:
      - Lakera Guard — runtime AI security screening for LLM applications with sub-50ms p95 latency
      - Single-endpoint /v2/guard API following the OpenAI chat completions message format
      - Detectors for prompt attacks, data leakage, PII, content moderation, and unknown links
      - 100+ language coverage and multimodal / model-agnostic protection
      - L1–L5 confidence levels aligned with OWASP paranoia level conventions
      - Per-detector breakdown and character-level payload match locations for masking
      - /v2/guard/results endpoint for offline threshold tuning and quality monitoring
      - Policies and Projects as first-class concepts — central policy control without code changes
      - SaaS deployment with daily detector model updates and real-time analytics
      - Self-hosted deployment with bi-weekly model updates and Kubernetes probes
      - Regional endpoints — Global, US East, US West, EU West, Asia Pacific
      - Lakera Red — AI red-teaming with direct and indirect attack simulations and risk-based findings
      - Workforce AI Security — monitors employee AI usage across applications and browsers
      - AI Agent Security — runtime protection for autonomous agents and tool use
      - Gandalf — 1M+ player prompt injection challenge feeding detector training signal
      - PINT Benchmark — open prompt injection detection benchmark on GitHub
      - LiteLLM, AI gateway, and RAG pipeline integration patterns
      - MCP server at docs.lakera.ai/_mcp/server for AI client tools
      - Browser extension (Chrome) for ChatGPT data leak protection
      - Part of Check Point Software Technologies as of 2025
    sources:
      - https://www.lakera.ai/
      - https://docs.lakera.ai/docs/api
      - https://docs.lakera.ai/docs/api/guard
      - https://docs.lakera.ai/docs/api/results
      - https://docs.lakera.ai/docs/quickstart
      - https://docs.lakera.ai/guard
      - https://docs.lakera.ai/docs/integration
      - https://github.com/lakeraai
    updated: '2026-05-25'
created: '2026-05-25T00:00:00.000Z'
modified: '2026-05-25'
position: Consuming
description: >-
  Lakera is an AI-native security platform that protects GenAI applications, agents, and workforces from prompt
  injection, data leakage, PII exposure, content violations, and malicious links. Lakera Guard exposes a single-endpoint
  runtime screening API (/v2/guard) that accepts OpenAI-style chat messages and returns a flagged decision in sub-50ms;
  a companion /v2/guard/results endpoint returns L1–L5 detector confidence levels for offline analysis. The Lakera
  Platform API lets Enterprise customers manage policies (detector sensitivity and action) and projects (per-application
  bindings) programmatically. Founded in Zurich and acquired by Check Point Software in 2025, Lakera also operates
  Gandalf, the 1M+ player prompt injection challenge that feeds its detector training pipeline, and publishes the
  open-source PINT benchmark for prompt injection detection evaluation.
maintainers:
  - FN: Kin Lane
    email: info@apievangelist.com
    X: apievangelist
    url: https://apievangelist.com
specificationVersion: '0.16'