Continuous attacks on your voice AI agents and chat AI agents
SottoAI runs continuous attacks on your voice AI agents and chat AI agents: the same tricks a manipulative caller or a bored hacker would try, so you see what breaks before they do.
SottoAI owns smart attackers. Real conversations. Real pressure. Run continuously, so SottoAI finds the cracks in your AI agent before a real attacker does.
However you built it, Sotto tests it: agents on vendor platforms (ElevenLabs, PolyAI, Retell, Deepgram), agent platforms themselves, and in-house stacks (LiveKit, Pipecat, custom orchestration). No integration or source code needed to start.
Evals check if your agent gives the right answer. SottoAI checks if it can be broken.