Test Runtime Governance Without Your Own Agent

Pick the path that fits your team

Try sample agent scenarios

Each sample is a realistic planner output. Pick one and run it — you'll see the user task, the proposed trajectory, the exact governance request payload, the live verdict, and the resulting execution decision. Nothing is ever executed; the engine only inspects the proposed trajectory before runtime.

[
  {
    "tool": "retrieve_customer_records",
    "args": {
      "scope": "internal"
    }
  },
  {
    "tool": "summarize_findings",
    "args": {
      "audience": "internal"
    }
  }
]

{
  "trajectory": [
    {
      "tool": "retrieve_customer_records",
      "args": {
        "scope": "internal"
      }
    },
    {
      "tool": "summarize_findings",
      "args": {
        "audience": "internal"
      }
    }
  ]
}

Type a task — watch the planner and the live engine decide

Describe any task in plain English. The same Hugging Face planner proposes a tool trajectory, and that trajectory is sent to the live Runtime Governance engine. The verdict always comes from the engine — it is never hardcoded.

The planner is a small open-weight model and may under- or over-plan a task; the verdict is always evaluated on the trajectory it actually proposed. No tools are ever executed.

The middleware pattern, demonstrated end to end

You don't need a production agent to test the architecture

You do not need a production agent to test the architecture. The demo planner creates realistic tool trajectories so your team can see where Runtime Governance sits before deploying agentic workflows.

Your model, your endpoint, or our demo endpoint

curl -s https://resurrection-tech.com/api/evaluate-trajectory \
  -H 'content-type: application/json' \
  -d '{"trajectory":[
        {"tool":"retrieve_customer_records","args":{}},
        {"tool":"export_customer_data","args":{}},
        {"tool":"send_to_external_webhook","args":{"url":"https://partner.example/ingest"}}
      ]}'

From demo → assessment → pilot → integration

Testing without an agent is step one. When you're ready, the same engine powers a full assessment of your real tools, a scoped pilot, and production integration.