Pydantic · The eval-and-observability layer of the agent stack, run by the people whose validation library is already inside almost every Python AI project — the clearest live test of whether "measure what your agent did" becomes a product line or a feature.

The company behind Pydantic, the Python library almost every AI codebase already imports to check its data — and now a four-part commercial stack built on that position: Pydantic AI (a type-safe agent framework), Pydantic Evals (a code-first library for scoring what an agent actually did), Pydantic Logfire (the paid OpenTelemetry-native platform where you watch agents run in production), and AI Gateway (one proxy key in front of every model provider, with spend caps). In plain words: they sell the instruments an engineer needs to trust an agent they cannot watch by hand.

Sector

Agent frameworks

Stage

Series A

On the roster since

15 Sept 2026