AI Resources

Request Trace (Audit Trail)

A sequence diagram is how systems are actually read: actors across the top, time flowing down, every call explicit. Run the trace to watch one request move through the system. The log beneath it appends as it runs, because in this architecture, the audit trail is not a feature. It is the medium. Select any actor for depth.

trace: one-request.e2e · owned-mcp architecture
idlePress run, or select any actor.
personjudgmentorchestratorclaude · manageragentsclaude ×3 · rolesmcp-gatewayowned · auditedtruth (git)tokens.json · contractsknowledgerag · memory · datapipeline (ci)deterministicfigmacanvas · mirrorintent("outcome, in plain language")plan(): steps · owners · checkpointsloop [ until goal met ]delegate(task, role, constitution)request(scope), least privilegegateway: check scopes → write audit-logread: tokens.json · prop-contractscomponents · tokensretrieve: personas · research · decisions · evidence (rag)top-k context, citedcontextresults → synthesize · re-delegatesubmit(draft)enforce: disclosures · a11y · brand · perf-budgets✓ pass, or it stops herecheckpoint: request approvalapprove() · name + time recordedship: write canvasmerge: pr → gitBACKGROUND · CONTINUOUSsync (ci): tokens.json → style-dictionary → theme · → figma variableslog: outcomes · decisions → memoryaudit-log: gateway calls + pipeline events

Trace Log: Appends as the Request Runs · This Is the Audit Trail

// idle. run the trace, or select an actor above.

An illustrative run: the timings and entry counts are example values, not measurements.