Public demo

Show the gateway stopping AI risk before action.

I’m going to attack an AI agent live — and show exactly where Firefish stops it.

Firefish protects the full AI boundary: prompts, retrieved content, outputs, streaming, tools, and audit.

Conference mode

Five safe, deterministic scenarios.

Run safe AI red-team scenarios against Firefish. The public demo only runs prebuilt synthetic scenarios and never executes real tools.

Ready

Pick a scenario.

Firefish will show the verdict, reason codes, redacted trace, and stage-friendly flow.

Input detected
Risk scored
Policy applied
Action taken
Audit trace saved
Red-Team Lab

Run safe AI red-team scenarios against Firefish.

Replay prompt injection, hidden RAG injection, tool hijack, streaming, secret leakage, and benign hard-negative drills with synthetic data only.

Open Red-Team Lab
Conference flow

Prompt → RAG Content → Model Output → Streaming → Tool Calls → Audit Trail

The public demo keeps the story approachable while the protected app keeps operational controls behind dashboard auth when configured.

Launch Dashboard