Configuration languages shape how developers think about their systems. We chose TOML because it brings object-oriented thinking to configuration—concepts with operations, not nested metadata.
String matching won't work for non-deterministic systems. We built a library that scores semantic similarity as floats, allowing for natural variation in agent outputs.
AgentKit creates an entirely new category that democratizes AI automation for everyone who knows their workflows intimately. This is what we've been anticipating.
The parallels between traditional software testing and AI agent evaluation are undeniable. Anyone who tells you evaluations aren't necessary has very little experience in software development.
Test your agents like you test your code. Get automated agent testing running on every pull request in just 5 minutes.
Clean code principles applied to AI agents. Why evaluation should be infrastructure, not application logic.