Arcjet

Arcjet positions itself as runtime security for AI agents, moving enforcement from network gateways or prompt scanning to the actual action boundary where tools, APIs, and database calls execute. The website shows an SDK-first product that offers prompt injection detection, sensitive information detection, token/spend budgets, bot protection, rate limiting, and WAF capabilities across many web and AI agent frameworks. Its core argument is that identity cannot predict unsafe agent actions; every step must be observed, enforced, and audited inside the application. Revenue, growth, team size, and traction are not publicly disclosed from the provided content. The approach is technically ambitious and strategically differentiated, but the market is young and competitive.