RFB Guard (Real-time Validation)
Policy engine, risk scoring, multi-critic voting, and fail-closed braking for critical actions.
Risk Firewall Bridge for high-stakes automation
RFB sits as middleware/proxy between an AI agent and your target system. A critic-agent layer validates each transaction for policy, context, and integrity before execution.
RFB Guard
Realtime Verify API
Critic Layer
5 Critics + Fallbacks
RFB Analytics
Audit + KPI Dashboard
Transaction Lifecycle
- Agent sends intent + parameters to
POST /api/verify. - PolicyCritic + LLM Critics return strict JSON votes.
- Decision engine aggregates to approve/deny/needs_human.
- Tamper-evident audit hash chain signs the record.
- Optional connector execution report checks integrity mismatch.
RFB API Plus includes a contractual RFB Fault Event Guarantee with capped service credits, not insurance.
Critic Agent Layer
DeepSeek, Llama-4 Maverick, Pollinations Perplexity, and OpenAI-Large with timeout/fallback chaining.
RFB Analytics
Audit timeline, signed decision records, usage KPIs, claims tracking, and integration health.
No "all damages" promise
RFB API Plus only applies to RFB Fault Events: policy/decision integrity breach, broken audit chain on an APPROVED transaction, or connector mismatch between approved_action and executed_action.