CloudGrid

Observability is not a place to rent your own production truth.

CloudGrid is an open-source, self-hosted OpenTelemetry observability platform that keeps traces, logs, metrics, dashboards, and AI-agent evaluation inside infrastructure the enterprise operates.

cloudgrid.dev Repository

Enterprise pressure

Why it matters

Enterprises need observability, but telemetry often contains production behavior, customer data, prompts, completions, identifiers, and operational evidence. SaaS APM moves that truth outside the perimeter. Fragmented OSS stacks keep data closer but split traces, logs, metrics, dashboards, and AI evaluation into too many systems.

Product answer

What changes

CloudGrid treats observability as an owned control surface. OTLP data enters through a collector, crosses a message bridge, is persisted through storage services, and is queried through a project-scoped UI for traces, logs, metrics, dashboards, and AI-agent evaluation.

Operating model

How the model holds

Services and agents send OTLP signals into a collector; private services exchange work over NATS JetStream, persist through a SurrealDB adapter, and expose project-scoped traces, logs, metrics, dashboards, and AI-evaluation views through the BFF.

cloud topology CloudGrid

your perimeter

OTLP collector

traces logs metrics dashboards ai eval

adapter boundary NATS + SurrealDB signals stay project-scoped

Decision path

CloudGrid starts from a board-level problem: production truth is becoming too sensitive and too expensive to rent blindly.

Telemetry now includes customer behavior, identifiers, prompts, completions, agent traces, failure evidence, and operational audit material. Sending all of that outside the enterprise perimeter can become a cost issue, a legal issue, a procurement issue, and a trust issue at the same time.

CloudGrid is the open-source answer for teams that want to own the observability control plane. OpenTelemetry signals enter through a collector, move through explicit services, and become traces, logs, metrics, dashboards, and AI-agent evaluation in one project-scoped product surface.

It is not a compliance badge and it is not a managed APM clone. It is a practical self-hosted foundation for enterprises that want telemetry ownership before observability turns into another lock-in layer.

Enterprise relation

Where this connects to enterprise pressure.

Telemetry ownership supports legal, procurement, and data-residency discussions before observability becomes SaaS lock-in.

Sensitive traces, logs, metrics, prompts, completions, and agent evaluation evidence can stay inside operated infrastructure.

Auditability improves because production signals and AI-agent evaluation live in one project-scoped operational surface.

Cost and retention policies become architecture decisions the enterprise can inspect instead of vendor-side surprises.

Capabilities

CloudGrid turns pressure into an operating model.

OpenTelemetry-native ingest for traces, logs, metrics, and GenAI spans

One product surface for production signals and AI-agent evaluation

Self-hosted open-source platform with telemetry staying inside your network

Adapter ports for storage, message bridge, auth providers, and evaluation harnesses

v1 stack around SurrealDB, NATS JetStream, OIDC providers, and puristajs/harness

Enterprise use cases

Where decision makers should care.

When telemetry ownership is a requirement, not a preference

When agent evaluation needs to sit next to the spans that produced the run

When teams want one query and project model for traces, logs, metrics, and dashboards

When infrastructure dependencies should be explicit instead of welded into the product

Links

Website cloudgrid.dev Repository Repository

Related writing will appear here as the article collection grows.