kakunin-ai/kakunin-core — explained in plain English
Analysis updated 2026-05-18
Issue a verifiable identity certificate to an AI agent your system runs.
Automatically revoke an AI agent's access when its behavior becomes risky.
Generate compliance evidence for AI agent activity under regulations like the EU AI Act.
Integrate agent identity checks into a LangChain or similar agent framework.
| kakunin-ai/kakunin-core | cschanhniem/clawping | cub3y0nd/entropic | |
|---|---|---|---|
| Stars | 24 | 24 | 24 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | hard | moderate | moderate |
| Complexity | 5/5 | 3/5 | 3/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Full self-hosting requires provisioning around fifty environment variables across Supabase, AWS KMS, Upstash, Stripe, and more.
Kakunin is infrastructure for giving AI agents a verified identity and for keeping track of how they behave. It issues each agent a digital certificate, watches its behavior in real time, gives it a risk score, and automatically revokes that certificate if the score crosses a set threshold, cutting the agent off within seconds. It also produces records meant to satisfy regulations such as the EU's MiCA and AI Act rules. This particular repository is the platform itself, the control plane behind the hosted service at kakunin.ai. It includes the web application, the piece that talks to the certificate authority, the behavior scoring engine, the API, and the reporting pipeline that generates compliance records. The README is clear that most developers do not need to run this code themselves. Instead, they are pointed toward lightweight software development kits for TypeScript and Python that talk to the hosted service directly and come with free sandbox keys, plus integrations for common agent frameworks and a server for connecting through the Model Context Protocol. The reasoning given for keeping the certificate authority itself as a hosted service, rather than something anyone can run their own copy of, is that a trust system only works if every party can check a certificate against one shared authority. So while this repository's code can be run on its own, its security guarantees depend on that central hosted service. The platform is built with Next.js, TypeScript, Supabase for its database, AWS KMS for managing cryptographic keys, and several other external services for things like billing, email, and background jobs. Running it yourself means providing your own accounts and credentials for around fifty configuration variables covering all of these services, since the application itself does not store any state on its own. A Docker image is published for convenience, and a docker-compose file can spin up a local Postgres and Redis for development, though the README notes this still does not replace the full set of hosted services the platform depends on. The platform code itself is licensed under AGPL-3.0, while the client SDKs use the more permissive Apache-2.0 license.
A platform that issues digital certificates to AI agents, scores their behavior for risk, and automatically revokes access if an agent misbehaves, aimed at regulatory compliance.
Mainly TypeScript. The stack also includes TypeScript, Next.js, Supabase.
Setup difficulty is rated hard, with roughly 1day+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Don't trust strangers blindly. Verify against the repo.