I've been working on Kuberna Labs for a few months, and I'm excited to share the first public release. It's an open-source SDK and infrastructure that lets developers build AI agents capable of executing cross-chain transactions autonomously.
The Problem
Building an agent that can actually do something on-chain (swap, lend, monitor) requires managing private keys, estimating gas, handling retries, and orchestrating cross-chain flows. It's a huge distraction from agent logic.
What Kuberna Does
Provides a unified SDK (@kuberna/sdk) to create, deploy, and manage agents.
Uses an intent-based architecture (ERC-7683) where agents express desired outcomes, and a solver network competes to fulfill them.
Supports TEE deployment (Phala, Marlin) for secure, verifiable execution.
Integrates zkTLS (Reclaim, zkPass) to fetch verified Web2 data privately.
Includes escrow contracts to lock funds until tasks are completed.
Current Status
All core contracts are written and tested. The SDK is functional (see README for quickstart). We have a full Docker Compose setup for local development. Everything is open-source under the MIT license.
Looking for Feedback
I'd love your thoughts on the architecture, the SDK API, and the overall vision. If you're building in the AI x Web3 space, I'd especially appreciate your input.