Evaluate ingest → evidence → analysis → surfaces on a representative production boundary. Send telemetry directly to Epok, inspect evidence items and receipts, and check coverage before judging incident outcomes. Your team and agents use the same facts; current observability and paging systems remain the control during evaluation.
Evaluating beside Datadog? Ship to both for 14 days; your paging stays authoritative. Your shippers keep their existing output and gain a direct Epok destination. Compare recognition, recovery recommendations, root-cause evidence and forecasts against agreed incident windows before expanding the deployment.
shadow mode · independent telemetry path · current paging authoritative · no automated remediation
Choose a service group, ownership domain, environment, critical user journey, or incident class. Record approved signals and fields, expected volume, data classification, region, retention, notification destinations and the evidence your team and agents need to check.
Fan out approved telemetry directly to Epok’s owned ingest through OpenTelemetry or an open shipper using an ingest-scoped key. Give the destination its own queue, timeout, retry policy and volume ceiling. No proprietary Epok agent for server-side telemetry.
Existing telemetry destinations, monitors, dashboards, paging and incident command remain authoritative during evaluation. Assess recovery recommendations from change and blast radius while root-cause analysis continues. We recommend; you execute.
Use agreed incident windows, synthetic failures or approved failure exercises across related systems. Follow evidence IDs to receipts and inspect coverage, including what was not seen. Request MCP gateway access for your agents and test the same evidence from the app and the gateway.
Classify each incident as correct, incorrect, abstained, or missed. Measure alert fanout, time from first page to verified cause, false pages, telemetry coverage and responder effort. Assess recovery recommendations and forecast uncertainty against the incident evidence as separate outcomes.
Review the pre-agreed quality, security, and operational gates. Expand through a separately approved change, keep the bounded deployment, or revoke the key and remove the exporter for a clean exit.
Do not combine abstentions with wrong conclusions, and do not call a trial successful because data arrived. Agree on the incident cohort and thresholds before ingest begins, then review both systems against the same windows. Epok commits a cause only with supporting evidence. Inspect found, none in retained data and not run separately, along with partial or missing inputs. Follow receipts through the MCP agent gateway, with access on request, to evaluate agent investigations alongside your team's.
Google OAuth, scoped ingest/read API keys, EU-hosted primary infrastructure, DPA on request, and timestamped alert-action history.
Capacity above published self-service limits, custom retention or residency, dedicated or self-hosted deployment, and contracted support requirements.
Microsoft login, SAML enterprise SSO and SCIM are not supported; Epok does not hold a completed SOC 2 Type II report. Review these constraints against your organization's production approval requirements.
Inspect the product without signup, run a self-service workspace, or define a governed evaluation with us.