Built a tool to test AI APIs and chatbots before release
We’ve been building PromptBrake — a platform for testing AI APIs and chatbots before they go live.
The idea came from seeing more teams ship AI features without really knowing how those systems behave under adversarial or unexpected input.
PromptBrake runs real-world attack scenarios against the actual endpoint your app ships to help uncover things like:
prompt injection
jailbreaks
unsafe responses
data leakage
tool misuse
off-script chatbot behavior
Results come back as PASS, WARN, and FAIL findings with evidence and remediation guidance so teams can quickly reproduce and fix issues before production.
We also recently added:
Chatbot Launch Testing
Replay Packs based on real-world AI incidents
CI release gate support
Free security tools and checklists
Would genuinely love feedback from people building with AI:
Is the positioning clear?
Does this solve a real pain point?
What would make you trust/use a tool like this?
Replies