Lyrie

Lyrie.ai, built by Dubai-based OTT Cybersecurity LLC and led by founder Guy Sheetrit, describes itself as security infrastructure for the AI agent era. The company exited stealth in May 2026 with a $2 million pre-seed round and simultaneously released the Agent Trust Protocol (ATP), an open, MIT-licensed cryptographic standard for establishing AI agent identity, permission scope, attestation, delegation, and revocation, which it has submitted toward IETF standardization. The same announcement disclosed that Lyrie was accepted into the first batch of Anthropic’s Cyber Verification Program, an independently reportable credibility signal beyond the company’s own claims.

Lyrie’s open-source engine (lyrie-omega, a Python CLI, paired with a TypeScript ATP SDK) runs a seven-phase autonomous pentesting pipeline, covering reconnaissance, fingerprinting, scanning, exploitation, proof-of-concept generation, and reporting, against live targets and source code, outputting SARIF-format results compatible with GitHub code scanning, plus an adversarial red-teaming module for LLM endpoints. Separately, the company’s commercial site markets a much broader consolidated suite (WAF, endpoint defense/EDR, vulnerability scanning, automated patching, breach monitoring, and free CAPTCHA) pitched as a single-platform replacement for a multi-vendor security stack.

The combination of a genuinely novel open protocol, a working open-source tool with real GitHub traction (300+ stars), and coverage in trade outlets like Help Net Security and CIO.com supports that Lyrie is a real, operating company. But it is very early stage, pre-seed, small team, no disclosed enterprise customer base or independent efficacy testing, and its ‘one platform replaces your entire security stack’ marketing significantly outpaces what a startup at this stage can be expected to have proven.