Type-safe LLM outputs with Zod: stop guessing what the model returns.

I shipped a classifier to production in January. The prompt asked for JSON with a single category field. For three weeks it worked fine. Then the model started returning {"category":"bug","explanation":"this looks like a crash"} and the consumer threw a runtime error because it only expected one key. No schema change, no deploy. The model just decided to be helpful.

Zod plus a bit of discipline around the parse step closes that gap. This tutorial walks through defining schemas for LLM output shapes, using them with the Vercel AI SDK and the raw Anthropic SDK, and building a retry loop that handles the cases where the model still gets it wrong.

TL;DR

Step

Type-safe LLM outputs with Zod: stop guessing what the model returns.

Related reading

Structured Output From Local LLMs: JSON That Never Breaks (Ollama + Zod)

Never trust an LLM's output directly. Here's the validation layer I put on…

Stop Begging Your LLM for Valid JSON: Self-Correcting Structured Output in…

Structured Outputs: How We Stopped Parsing LLM Responses by Hand

Trusting an LLM's JSON in production

Catch LLM Schema Drift Before It Breaks Production