Deterministic step budget · 2 MB scored in ~3.4 ms · 947-byte edge bundle, no database · Blue/green rollback
See how it compares to ODMDecision observability
See which rules matched and in what order, how the engine spent its step budget, whether the warehouse writer is healthy, how services connect, and whether error budgets are burning — from real Decision Server and Manager surfaces. We label Live, Partial, and Residual so residual causal AI or multi-region synthetic fleets are never sold as shipping product.
Live = product APIs and consoles today. Partial = limited or operator-assisted. Residual = not invented on this page.
No credit card · Trial does not auto-convert · Export anytime
These capabilities ship against warehouse, engine telemetry, topology, SLO, and OTEL exporters — the decision-observability honesty ledger, not a generic APM storyboard.
Every execution is written asynchronously with zero hot-path overhead. Browse, filter, export, and inspect writer lag and queue depth — the same store KPI dashboards and replay audits read.
Ordered firings with fire order, steps, and residual_rules honesty on the engine strip. Last-writer field attribution when outputs are present — not a reconstructed code-level trace tree.
Selection rates, step cost, expensive-rule views, and fire frequency from engine probes and monitoring APIs — rule-level signal that host APM cannot invent.
Admission, shed, budgets, inflight, and capacity on health endpoints and the engine plane — so saturation is visible before latency graphs alone tell the story.
Live service/artifact topology and blast-radius impact from Manager topology APIs. Version-aware subjects and deep links — not a host process injection graph.
Server SLO panels, Prometheus burn dashboards, and error-budget tracking against decision health — business SLO continuous golden monitors stay Partial/Residual where not productized.
Real dm.* semantic conventions, resource attributes, and metrics from exporters — only attributes that exist in code. No fictional span catalog or browser OTLP waterfall claim.
Enterprise buyers ask about causal AI-class AI and multi-region synthetic fleets. We answer with Residual labels instead of inventing product surfaces.
causal AI-class causal AI
Residual — Not productized yet — stated for honesty, not marketed as readyNo automatic causal graph, predicted business-impact AI, or causal APM AI clone. Grounded explain-decision / RCA narrates recorded decisions only where those Manager routes exist.
Multi-region synthetic fleet
Residual — Not productized yet — stated for honesty, not marketed as readyCLI and pipeline suite runs are Live stand-ins. Continuous multi-region synthetic monitors as a product fleet remain Residual — not a silent claim of global mesh probes.
APM topology / code-trace clones
Residual — Not productized yet — stated for honesty, not marketed as readyNo auto-discovered host topology process map, no code-level trace tree from matched[] alone. Topology is decision-service and artifact oriented with residual honesty for full APM overlays.
Async KPI writing, writer health, and regulatory replay audits — the store behind every warehouse trace.
Warehouse feature pageDeploy, execute, measure — runtime console with warehouse, SLO, topology, and engine surfaces.
Platform · Decision ServerSample decisions and illustrative traces in the browser — no account required.
Run the live demoTrial includes Decision Server observability surfaces. Prefer a guided walkthrough of warehouse traces, topology, and SLOs? Book time with an executive seller.
No credit card · Trial does not auto-convert · Export anytime