Skip to main content

​
What is Prompt Inspector?

Prompt Inspector is an open-source security layer designed to protect your LLM-powered applications and AI agents from prompt injection attacks, sensitive content leaks, and other adversarial inputs. It acts as a guardrails between untrusted user input and your language model, analyzing every prompt in real time before it reaches your AI system.

Quick Start

Get up and running in under 5 minutes with our API, SDK or MCP.

Detection Architecture

Learn how the multi-layer detection engine works.

Python SDK

Integrate with Python in just a few lines of code.

Node.js SDK

Add prompt safety to your TypeScript or JavaScript app.

MCP Integration

Integrate with Claude Desktop and other MCP-compatible AI tools.

More Coming Soon

Additional features and integrations are on the way. Stay tuned!

​
Why Prompt Inspector?

Real-time Detection

Multi-layer analysis pipeline catches threats that single-model approaches miss, without compromising on speed.

Sensitive Word Filtering

Per-tenant custom sensitive word lists with support for literal matching and regex patterns.

Enterprise Payload Library

Million-scale attack payload database with real-time updates, continuously refined from production threats.

​
How It Works

Prompt Inspector uses a multi-layer funnel architecture to efficiently detect threats at different levels of complexity:
  1. Custom Sensitive Words — Per-tenant keyword and regex matching using high-performance automaton algorithms
  2. Semantic Analysis — Deep vector-based detection model that understands meaning, not just keywords
  3. AI Review — Optional intelligent review for ambiguous cases in the gray zone
  4. Arbitration — Final decision aggregation and confidence scoring
  5. Self-Iterating Engine — Continuously learns from attack patterns to enrich the payload library and enhance detection capabilities
Each layer acts as a progressively finer filter, ensuring threats are caught early while keeping processing efficient.

​
Integration Options

REST API

Call the detection endpoint directly from any language or platform via HTTP.

Python SDK

Install via pip install prompt-inspector and start detecting in seconds.

Node.js SDK

Install via npm install prompt-inspector for TypeScript and JavaScript projects.

MCP Server

Connect AI coding tools like Cursor, VS Code Copilot, and Claude Desktop via Model Context Protocol.

​
Open Source

Prompt Inspector is open-source under the AGPL-3.0 license. Contributions, issues, and feature requests are welcome.

View on GitHub

Star the repo, report issues, or contribute to the project.

​
Enterprise Edition

We offer an enterprise version with additional services including payload library maintenance, customization, private deployment, and dedicated support. For commercial use and enterprise inquiries, contact us at [email protected].

Contact Sales

Get in touch for enterprise pricing, custom solutions, and dedicated support.