Turn every signal into owned operational evidence.
CloudGrid is the OpenTelemetry-native platform for teams that need traces, logs, metrics, dashboards, alerts, and AI evaluation to work from one project-scoped record. It is self-hosted, source-available, and built so telemetry stays inside the environment you control.
One self-hosted system. Many signals. One evidence model.
A leadership team does not need a separate answer for every telemetry format. It needs a reliable way to understand what happened, why it matters, who owns it, and which decision follows. CloudGrid keeps the signals distinct where they need precision and connected where the business needs evidence.
Explain the request path and the service boundary where behavior changed.
Make operational text searchable without losing trace and span context.
Track health, latency, volume, runtime, token, and cost signals over time.
Keep recurring operational questions visible for reviews and daily work.
Turn live and persisted telemetry into managed action with state and history.
Convert production examples into datasets, runs, comparisons, and promotion evidence.
Adapt brand, packaging, storage, bridge, auth, delivery, and execution boundaries.
From incident to decision, the evidence stays connected.
Enterprise teams rarely fail because one signal is missing. They lose time when the trace, log line, metric trend, dashboard, alert, dataset row, and evaluation result live in different products with different ownership rules. CloudGrid keeps those pieces in one project workspace and lets the right service own each data path.
CloudGrid receives OpenTelemetry traces, logs, and metrics into a project boundary. The same project also owns dashboards, alert rules, datasets, evaluation runs, and optimization results.
A log can point to a trace. A metric exemplar can point to a trace. A failed AI example can keep the source trace beside the dataset row and the evaluation result.
Platform teams investigate incidents, product teams review service behavior, and AI teams validate model or agent changes without exporting evidence into separate tools first.
The message bridge and adapter boundaries let ingestion, persistence, reads, alert delivery, and white-label customization evolve without turning the public UI into the integration layer.
Wired like a workflow. Audited like infrastructure.
CloudGrid is modular without asking users to operate the product as a pile of separate tools. The public UI talks to the BFF. Private services own telemetry semantics. The message bridge gives each path a contract, so scaling and customization stay controlled.
Collectors receive OTLP and publish bounded work. Public services do not become storage clients.
Request/reply reads, durable ingest, live fanout, and delivery dispatch move through explicit message contracts.
Write paths mutate storage. Read paths own filtering, sorting, grouping, counts, live matching, and authorization preparation.
Users see project evidence, dashboards, alerts, and evaluation decisions instead of infrastructure wiring.
Brand it, package it, extend it.
CloudGrid can fit your product portfolio, customer environment, or internal platform. The important point is not that every layer can be changed; it is that the layers are named, bounded, and documented so customization does not damage the operating model.
Evaluate agents next to the spans they came from.
CloudGrid turns the telemetry you already collect into AI evaluation evidence. Curate datasets from known examples or trace-derived failures, run evaluations against a target, compare expected and actual outputs, optimize prompts or examples, and promote the candidate with validation evidence the team can inspect.
- ✓ Schema-backed datasets with input, expected output, reason, split, and curation state.
- ✓ Per-row results include metrics, trajectory summaries, and links back to trace evidence.
- ✓ Provider profiles and model aliases stay project-scoped and controlled by settings.
Built in the open. Run on your own terms.
Source-available under Apache 2.0 with Commons Clause. One reviewable distribution, visible commercial paths, and telemetry that stays in your network unless you wire it to leave.