Axiom Documentation - Observability Platform

A monitor is a background task that periodically runs a query you define. For example, it can count the number of error messages in your logs over the previous 5 minutes. When the query results meet certain conditions, Axiom sends you an alert via a notifier.

Axiom operationalizes timestamped event data, which are structured key-value records that capture every digital interaction between humans, sensors, and machines. Event data is the atomic unit of activity for organizations and consists of: a timestamp indicating when the event occurred, and attributes (key-value pairs) providing context about the event.

A notifier defines how Axiom notifies you about monitor output. For example, Axiom can send you an email when a monitor triggers an alert. Notifiers handle the delivery of alerts.

Issues are created automatically from the Review workflow. When a reviewer inspects a conversation and submits negative feedback, Axiom attempts to match it to an existing issue. If no match is found, Axiom creates a new issue and adds the reviewed conversation as supporting evidence. The Issues page collects these recurring findings into a shared backlog for centralized management.

Logs serve dual compliance purposes: they satisfy specific mandated controls (retention periods, event review) and provide evidence for auditors that other controls are functioning. Log-based reporting, while not directly mandated by any framework, forms a critical practice for demonstrating due care across software development, access controls, networking, and other IT areas under CISO purview.

To view your billing history, go to Settings > Billing and click Billing history in the Current breakdown section. This shows details about your last bill and projected costs for your next bill.

Axiom will not collect or process sensitive personal information (as defined by the EU-US DPF) — including information about health, racial or ethnic origin, political opinions, religious beliefs, trade union membership, or sexual orientation — without first obtaining explicit opt-in consent from the individual.

Axiom introduces three new chart types—Filter Bar, Log Stream, and Scatter—to enhance visualization and analysis of logs, metrics, and traces. These complement existing charts like Time Series, Statistic, and Table, enabling users to pivot, drill down, and broaden their view of infrastructure data.

California residents have the right to request information about data collected in the past 12 months, including: categories of personal data collected, sources, business purpose, third parties shared with, and specific pieces of data collected. They also have the right to request deletion of their personal data. Contact [email protected] to exercise these rights. Note: If Axiom processes your data as a service provider for a customer, contact the customer directly first.

MetricsDB is a dedicated metrics datastore engineered for high-cardinality time-series data. Unlike traditional solutions that penalize dimensional complexity, MetricsDB embraces high-cardinality tags as a design principle. Features: native high-cardinality support without performance degradation or cost penalties; optimized storage format for time-series workloads with efficient compression and fast aggregations across millions of unique tag combinations.

Evaluation is the systematic process of measuring AI performance. AI systems fail in non-deterministic ways—the same prompt can produce different results, and edge cases emerge unpredictably. As systems grow from simple single-turn interactions to complex multi-agent systems, manual testing becomes unscalable. Systematic evaluation addresses this by: establishing baselines (measuring current performance before changes), preventing regressions (catching quality degradation before production), enabling experimentation (comparing models, prompts, or architectures), and building confidence (deploying changes knowing they improve aggregate performance).