Skip to content
O/THE OPEN ALTERNATIVE
↑↓to navigate↵to selectescto close
Grepture

Grepture

Trace, evaluate, and protect every LLM call with a drop-in TypeScript SDK. Features PII redaction, cost tracking, and EU-hosted infrastructure for AI-powered apps.

Grepture preview

Grepture is a privacy-focused AI gateway that sits directly on the request path to trace, evaluate, and protect every LLM call your application makes. Designed for teams shipping AI features, it provides full observability, built-in evals, and advanced PII redaction without storing your data.

Key Features

  • Hot-path proxy: Processes every request in-flight for PII redaction, threat blocking, and rule enforcement before data leaves your network
  • Full observability: Detailed request logs, multi-step traces with waterfall views, and full-text search across prompts and responses
  • Built-in evals: LLM-as-judge scoring runs on live traffic with six templates out of the box
  • Cost tracking: Real-time token cost tracking with per-model breakdowns
  • Multi-model gateway: Works with OpenAI, Anthropic, Gemini, Cohere, Mistral, AWS Bedrock, Hugging Face, Groq, Replicate, and more
  • PII redaction: 80+ detection patterns with reversible tokenized redaction (mask & restore with TTL)
  • Prompt injection protection: Detects and blocks adversarial inputs
  • Zero-data mode: Processes requests without writing content to disk; only operational metadata is logged
  • EU-hosted infrastructure: All services run in Frankfurt & Nuremberg with GDPR compliance
  • Developer-first SDK: Drop-in TypeScript SDK with zero dependencies, CLI for local dev, and pre-commit scanning

Tech / Self-Hosting

Grepture is open-source and can be self-hosted. The official repository provides the proxy implementation, SDK, and documentation for deployment. The cloud version is hosted in EU data centers (Germany & Ireland) with zero US data transfers. The system is designed for high availability with 77.93% uptime over 30 days.