| Website | useinstructor.com |
| Category | AI / Developer Tools |
| License | MIT (open source) |
| Pricing | Free (open source) |
Overview
Instructor is a Python library for extracting structured data from LLMs. Pydantic validation, automatic retries, streaming. Works with OpenAI, Anthropic, and more.
Pros
- MIT license
- Pydantic model validation
- Automatic retry on validation failure
- Streaming support
- Multi-provider (OpenAI, Anthropic)
- Simple API
Cons
- Python only
- Requires Pydantic knowledge
- Some models extract better than others
- Complex nested schemas can fail
- Retry adds latency
Verdict
Instructor turns LLMs into reliable structured data extractors. Define a Pydantic model, pass it to Instructor, and get validated, typed output from any LLM. When the model hallucinates or returns invalid data, Instructor automatically retries with the validation error. For any application that needs structured data from LLMs — not just text — Instructor is essential.