cardinalhq.io

What It Is

Cardinal positions itself as “the first observability data lake for Agents.” Its core idea is not a traditional SaaS monitoring platform, but writing full-fidelity telemetry data—logs, metrics, traces, events, and more—into the user’s own object storage. Lakerunner then indexes that data in place, while the Agent Runtime lets Claude, OpenAI, Gemini, Cursor, or custom Agents perform natural-language investigations and automated troubleshooting.

Core Capabilities

Cardinal emphasizes native OpenTelemetry ingestion: standard OTEL collectors send logs, metrics, and traces to object storage, while Cardinal performs indexing and querying inside the customer’s VPC. Its Agent-oriented capabilities include compound skills such as Outlier Detector, Correlation Finder, and Error Summarizer, which can automatically break down metric spikes, correlate releases/configuration/scaling events, and cluster large volumes of errors into reasoning-friendly summaries. Compared with a simple API wrapper, Cardinal focuses more on providing Agents with structured, evidence-backed troubleshooting results.

Pricing and Deployment

Cardinal explicitly states that “you pay for your cloud, not our cloud.” Its model is software licensing plus a support contract: no SaaS, no per-query billing, and no data egress. The software license covers all signal types, unlimited retention, unlimited queries, and updates during the license term. The support contract includes deployment, upgrades, capacity planning, and incident support. Users still need to pay for object storage, PostgreSQL, Kubernetes, and LLM token costs from services such as Bedrock, Vertex, and Azure OpenAI. Specific license pricing has not been publicly disclosed.

Pros and Cons

The main advantages are that data does not leave the customer’s cloud environment, making it suitable for teams sensitive to security, compliance, and cost control. The absence of query limits also makes it better suited to high-frequency, deep exploratory queries by Agents. Native OTEL support lowers the barrier to adoption. The downsides are that teams need operational expertise in Kubernetes, object storage, databases, and cloud LLM services. Public materials also lack complete API/SDK documentation, deployment guides, and pricing tables, so a thorough PoC is needed before procurement.

Who It’s For and Access from China

Cardinal is better suited to mid-to-large engineering teams, platform engineering teams, SREs, and enterprises that want to build or own their observability infrastructure. It is less suitable for small teams looking for an out-of-the-box product. Access from mainland China cannot be determined from the available content, and payment methods have not been disclosed. If access or compliance is restricted, alternatives to evaluate include Grafana Loki/Tempo/Mimir, Elastic Observability, Datadog, New Relic, Honeycomb, or a self-hosted ClickHouse + OpenTelemetry setup.