# Validator Master

> Provider-neutral behavioral evaluation infrastructure that applies the versioned Heartcode Protocol to observable AI responses and returns structured evidence with explicit scope and validity limitations.

- Public demo: https://demo.heartcodeprotocol.org/
- Heartcode Protocol: https://github.com/heartcodeprotocol/heartcode-protocol
- OpenAPI: https://api.heartcodeprotocol.org/openapi.json
- Security: https://api.heartcodeprotocol.org/.well-known/security.txt

## Canonical documentation

- [Validator Master](https://api.heartcodeprotocol.org/docs/validator-master): Provider-neutral behavioral evaluation infrastructure that applies the versioned Heartcode Protocol to observable AI responses and returns structured evidence.
- [Validator Master Quickstart](https://api.heartcodeprotocol.org/docs/quickstart): Minimal agent-oriented path for evaluating an existing response or invoking the bounded public provider-backed evaluation API.
- [Heartcode Methodology and Evidence](https://api.heartcodeprotocol.org/docs/methodology): How Validator Master evaluates observable response behavior and what the resulting evidence can and cannot establish.
- [Validator Master Public API](https://api.heartcodeprotocol.org/docs/api): Machine-readable overview of the current public evaluation endpoints and their operational boundaries.
- [Validator Master Limitations, Privacy, and Claim Boundaries](https://api.heartcodeprotocol.org/docs/limitations): Explicit limits on what Validator Master evidence establishes and how public-alpha requests should be handled.
- [Validator Master MCP Status](https://api.heartcodeprotocol.org/docs/mcp): Current Model Context Protocol capability and the boundary between local stdio tooling and future hosted agent transport.

## Current contracts

- Direct response evaluation: 1.0.0; scope response_text_only
- Provider-backed public evaluation: 1.0.0; providers openai, ollama, gemini, anthropic, xai, nvidia
- Arena: 1.0.0; 2-5 providers from openai, anthropic, google, xai, nvidia

## Important boundary

Scores are scoped behavioral-evaluation evidence. They do not prove universal correctness, factual truth, legal compliance, or total AI safety.