What is real?
Hosted-model calls, AutoGen agent execution, HTTP tool boundaries, independent policy evaluation, allowed SQLite mutation, blocked writes, audit records, OpenInference instrumentation, OTLP export, and Konfer processing.
Choose an execution
One workflow. Four governed outcomes.
Select a scenario, press play, or scrub the timeline. Every visual is driven by captured telemetry rather than a recreated result.
00:00 / 00:00
0of 0 captured spans
0from execution identities
0content redacted
0causally attributed
ReadyDiscover → Govern → Assure
Observed agent interaction graph
Nodes and handoffs appear only when supporting spans have arrived.
Press play to begin
Konfer will reconstruct the workflow from the captured OTLP spans.
Governance decision
Observed behavior and enterprise treatment remain distinct.
Pending
The independent policy gateway has not evaluated an action yet.
Live evidence stream
Select an event to inspect its redacted OTLP attributes.
Post-execution assurance
The completed trace connects identity, policy, intervention, and final state.
Not complete
Agent and handoff evidence accumulates during replay.
Pending
The enforcement decision will remain linked to its source spans.
Pending
The state comparison is finalized after enforcement.
Current redacted OTLP span
No prompts, model responses, tool inputs, tool outputs, or credentials are exposed.
{}
What is synthetic?
Vendor names, invoice records, approvals, and user requests are fictitious and explicitly marked as synthetic demo data. The replay makes no claim that the local ERP is Oracle, SAP, or another commercial platform.