Run PydanticAI agents on a governed runtime — Kohala
Typed agents, now with typed governance. Import your PydanticAI agent or wrap it with the kohala SDK — Pydantic output models keep validating while Kohala adds charters, caps, metering, and traces.
Explore Kohala
Frequently Asked Questions
- Can I run a PydanticAI agent on Kohala?
- Yes. Paste your PydanticAI code into the import wizard and your loop becomes the agent's scripts on the governed runtime, under a recorded charter, a server-enforced allowlist on the Kohala tools it can call, and a billing cap — unscheduled until you add a cadence. Or wrap the agent with the kohala Python SDK on your own infrastructure and report each run, metered against the same caps and checked by the same output validators. Your Pydantic output models keep validating results either way.
- How do Pydantic output models interact with Kohala validators?
- They stack. Your Pydantic model validates the shape of the output as usual, and Kohala's output validators enforce account-level policy — like content rules or required fields — before a run is marked complete.
- Does the type safety survive the wrap?
- Yes. The wrapped agent returns the same typed output your Pydantic model defines, plus run metadata: trace ID, credit cost, and latency.