kakunin-ai/kakunin-sdk-typescript

Kakunin is a TypeScript toolkit for giving AI agents a verifiable identity, aimed at companies that need to prove which AI agent did what, especially for financial or regulatory compliance reasons such as the EU AI Act. It works by talking to a hosted Kakunin service that issues each AI agent a real digital certificate, the same kind of technology, called X.509, used to secure websites. With this SDK you register an agent, get it a certificate, and then send in behavior events, such as an agent initiating a financial transaction, to receive a real time risk score back. Anyone can verify whether a given certificate is still active, revoked, or expired, without needing to log in. There is a free sandbox mode using special test keys, which issues real certificates from a test authority that have no legal or regulatory standing, useful for trying things out before using it for real. The toolkit also ships middleware you can drop into a service that receives requests from Kakunin certified agents, so that service can check for a valid certificate header before trusting the request. This middleware works with several popular web frameworks, keeps a short cache of certificates to avoid repeated lookups, and by default blocks a request rather than allowing it through when something cannot be verified. There is also a specific integration for Supabase, a hosted database service, that ties a verified agent's certificate to database permissions so an agent's queries are automatically restricted to only the rows it is allowed to see. An additional feature lets a service confirm that an incoming webhook message genuinely came from Kakunin, useful for reacting immediately when a certificate gets revoked. Getting started requires Node.js version 18 or newer, and the project is released under the Apache 2.0 license, a permissive license that allows commercial use as long as you keep the required notices.