# CrocoTiger

> CrocoTiger is a semantic fencing platform that adds a real-time security layer to LLM applications: it validates prompts, prevents prompt injection, and enforces strict contextual compliance.

CrocoTiger lets teams define a "semantic fence" for an AI model from a topic, a URL, or uploaded documents, then hardens and stress-tests it against a library of adversarial attack patterns (prompt injection, jailbreaks, red teaming, policy evasion, obfuscation) drawn from sources such as Garak, promptfoo, and IBM Research ATTQ. It ships as a web dashboard, a REST API, a Python SDK on PyPI, and an MCP server for connecting AI agents (e.g. Claude, Cursor) directly to the security fence. CrocoTiger is a product by TekDatum.

## Docs
- [Documentation](https://crocotiger.com/docs): Guide to the CrocoTiger dashboard — projects, builds, testing, monitoring, and settings.
- [Get Started](https://crocotiger.com/get-started): Deploying CrocoTiger via the AWS Marketplace listing.
- [Development](https://crocotiger.com/development): API integration, the Python SDK, the MCP server, and the Web UI.

## Product
- [Features](https://crocotiger.com/features): Project creation, project insights, attack simulation, build transparency, and the interactive playground.
- [Use Cases](https://crocotiger.com/use-cases): Example scenarios showing the semantic fence accepting and rejecting prompts.

## Optional
- [Contact](https://crocotiger.com/contact): Contact the CrocoTiger team.
- [Python SDK on PyPI](https://pypi.org/project/crocotiger-sdk/): Official crocotiger-sdk package.