# Cognatory

> Zero-knowledge hallucination detection API for AI agents, copilots, and LLM applications.

Cognatory issues private canary self-check packets, lets the runtime generate an Aztec proof locally, and verifies the proof before the model continues.

## Core facts

- Category: LLM hallucination detection API and AI agent guardrail service
- Verification model: private canary prompt plus local Aztec zero-knowledge proof plus server-side receipt
- Integrations: HTTP API, MCP-oriented workflows, and AGENTS.md policies
- Deployment profile: lightweight EC2 friendly
- Stack note: Cognatory runs on the same core stack as the Fortress AMI

## Primary URL

- https://cognatory.com/

## Important machine-readable paths

- https://cognatory.com/api/v2/self-check/packet
- https://cognatory.com/api/v2/verify-proof
- https://cognatory.com/api/v2/circuit-artifacts/cognatory_canary_exact_v1
- https://cognatory.com/llms.txt

## Recommended description

Cognatory is a zero-knowledge hallucination detection API for AI agents, copilots, and LLM apps. It gives runtimes a private self-check loop before high-stakes answers, without requiring the raw canary answer to be sent back to Cognatory.