Github - Recipes & examples

Community

Blog

Metatext

API reference

Documentation

Recipes

Dashboard

Support

Red Team Test Probes

Response Testing-List Test Probes

List Test Probes

Start securing your generative AI application in under 5 minutes

Quickstart

Protect every Generative AI application you build with Metatext AI

Introduction

Evaluate

Learn how to set up and manage policies and guardrails for your AI application

Policies and guardrails

Policy Rules Catalog

Jailbreak

Classifier

Factuality

Similarity

Rubric

Regex

Contains (string match) rule for guardrails

Contains (string match)

Model based rules

Policy example to block harmful content using classifier rule

Protect brand reputation by blocking harmful content

Mensure and detect uncertainty in AI-generated content

Detect hallucinations with uncertainty

Policy example to blocklist names, keywords, or other sensitive information

Blocklist

Policy example to ensure complex business guidelines using rubric rule

Ensure complex business guidelines

Policy example to block off-topic responses using classifier rule

Block off-topic responses

Policy example to detect direct prompt injection using jailbreak rule

Detect direct prompt injection

Policy example to detect PII exposure using pii rule

Detect PII exposure

Available probes for testing AI vulnerabilities

Probes Catalog

Automated red team tests to scan vulnerabilities

Scan vulnerabilities

Application setup

Directly access a subset of our backend endpoints through HTTPS.

Authentication

Create an application

Delete an application

Create a new policy for a specific application.

Create a policy / guardrails

Update an existing policy for a specific application.

Update a policy and guardrails

List all policies for a specific application.

List policies and guardrails

Delete a policy by ID for a specific application.

Delete a policy and guardrails

Red Team Test Vulnerabilities

You can directly access a subset of our inference endpoints through HTTPS.

Classify

Extract

Generate

OpenAI compatible

This endpoint allows you to fine-tune a model within a specified project.

AI Safety endpoints

Finetune & Inference endpoints

Red Team Test Probes

Response