Policy before execution
Actions pass through an explicit execution plan, dry-run and policy authorization gate before the provider can execute.
A real AI agent requests an external action. RBEK decides whether it is authorized, executes the third-party API call through a governed boundary, and produces verifiable proof of what happened.
Try before the live example
The default developer proof requires no API key and no model inference. RBEK policy enforcement and RBEK evidence are real.
✓ Deterministic Execution Verifiedcurl -fsSL https://releases.rbekplatform.com/cli/stable/install.sh | bash && \
curl -fsSL https://raw.githubusercontent.com/rbekplatform/rbek/main/examples/real-governed-agent/demo.sh | bash
Zero-key mode: model inference not required · external governed action network execution: NO · policy enforcement: REAL · evidence: REAL.
Then continue below to the live example: real OpenAI model inference, a real Open-Meteo HTTPS action, receipt and certification.
The AI model does not call the external provider directly. Execution authority remains inside the RBEK governed execution boundary.
Agent: OpenAI / gpt-5-mini Action: Get current weather Third-party API: api.open-meteo.com Execution mode: external-controlled Policy authorization PASS HTTPS execution PASS Network execution TRUE External API execution TRUE Database execution FALSE Receipt VALID Certification VALID Agent response: Temperature: 25.9°C. RBEK receipt valid: yes. RBEK certification valid: yes.
Models can reason and request actions. RBEK controls whether and how those actions execute.
Actions pass through an explicit execution plan, dry-run and policy authorization gate before the provider can execute.
The external call is performed by RBEK through the authorized provider boundary — not directly by the AI agent.
RBEK records what executed and produces a verifiable receipt and promoted execution certification.
Open-Meteo is used here because the execution is easy to verify publicly. The governed execution model is designed for real business systems.
RBEK is the governed execution layer for AI agents, workflows and software.