Quick Start

EvalDesk is an AI-native, expert-verified evaluation platform. An AI judge scores every agent answer; a credentialed expert verifies the uncertain ones; you get a cryptographically signed certificate an auditor accepts.

1. Sign up

Create an account with your email. No credit card required.

2. Create a project

A project represents one AI agent. Give it a name and your agent's HTTP endpoint URL.

3. Add test cases

Each test case has an input and an expected output. Optionally add a category (for compliance) or context (for RAG faithfulness).

4. Run eval

The AI judge scores every answer. Uncertain or disputed cases route to human review.

5. Review + sign off

A domain expert reviews flagged cases (blind, keyboard-first). On sign-off, EvalDesk issues an Ed25519-signed certificate with the compliance coverage matrix.