Observability Simulated 4 MCP tools

Observability Engine

Logs + metrics viewer with anomaly detection animation.

How you use it

When adding a service path, observe_plan early. Attach evidence before release_enforce on paths that require it.

Token savings

Blind debugging without logs is pure token waste. Requiring telemetry up front shortens incidents.

Consistency

Governed services carry the same log/metric obligations.

In action

Observability

Contract

Logs + metrics required for service X.

Scan

Plan

Enforce

Interactive demo

Telemetry Idle

14:02:01 commit_checkpoint ok scope=sdk

14:02:03 confidence_evaluate score=0.81

14:02:05 mcp_call latency=142ms

14:02:08 anomaly: latency spike 890ms

14:02:09 sentinel reflex triggered

MCP calls/min 42

Avg latency 890ms

Error rate 0.2%

Simulated behavior only — does not execute proprietary AGS engine code.

Diagram

Example usage

# MCP: observe_enforce
{}

# MCP: observe_plan
{ "serviceId": "ags-commerce" }