Live trust stack
The model is the last thing we trust.
Send a request through Pramagent. Deterministic layers decide before the provider is ever called, a second model judges what comes back, and the whole run is sealed into a hash chain you can verify. Watch where each request stops.
1Choose a request
2Or write your own
3Active guards
PII scrubbing
Injection guard
Safety rules
Output judge
Human approval (HITL)
Connection - bring your own key
NVIDIA NIM keys use the selected model; OpenAI keys use gpt-4o-mini. Used only for this request. The demo never persists the key.
The trust stack
readySafe output
prev
this
pip install pramagent
github.com/sriram7737/pramagent
Implementation status
The model is not the authority. Deterministic code outside it is.