PyRIT is a harness: a target to send prompts to, converters that rewrite them, scorers that judge what came back, and a memory database that keeps every exchange. It is for testing systems you are authorised to test.
Reach for it when
- You are responsible for an AI system and have to probe it
- You want every attempt recorded and queryable afterwards
- You need converters and scorers you can write yourself
Look elsewhere when
- You want a scanner with attacks already written. Look at Garak
- You want to test prompt quality rather than safety. That is an eval
What the lessons will cover
- 01A target you can run locally
- 02Memory, and what it records
- 03Attacks and objectives
- 04Scorers and verdicts
- 05Converters, alone and in an attack
- 06Writing your own
- 07Querying the memory
- 08Labels and a real database
Others in red teaming
Common questions
Is PyRIT free to use?
PyRIT is open source and free to run yourself. You still pay whichever model provider you point it at, and this tutorial is free with no signup.
Do I need to know Python to use PyRIT?
Basic Python is enough. Functions, dictionaries and imports cover most of what PyRIT asks of you.
When should I not use PyRIT?
You want a scanner with attacks already written. Look at Garak. You want to test prompt quality rather than safety. That is an eval.
Lessons for PyRIT are being written. Meanwhile the Agent Governance Toolkit tutorial covers the same ground: state, tools, loops, memory and human approval. Most of it carries straight over.
Start the Agent Governance Toolkit tutorial →