How agent-ready is Datadog?

Datadog's agentability audit reveals a score of 36/100 on its homepage, indicating significant friction for autonomous AI agents attempting to parse and interact with the platform. While machine readability scores well at 87/100, the site struggles with transparency (0/100), defaults (18/100), and shadow UI avoidance (15/100). These gaps mean agents face difficulty understanding system state, predicting outcomes, and operating controls reliably.

The pricing page (40/100) and documentation (25/100) show similar patterns. Strong HTML structure supports basic parsing, but the absence of transparency mechanisms—such as confidence indicators, audit logs, or source citations—leaves agents unable to validate outputs or trace decisions. Combined with weak defaults and control patterns, this creates an environment where agents must rely heavily on trial-and-error rather than predictable interaction patterns.

Improving agent-readiness will require addressing transparency fundamentals and strengthening interactive affordances. The technical foundation for machine parsing exists; the challenge lies in surfacing system reasoning and providing clearer operational guardrails.

To improve agentability, Datadog should prioritize the following concrete changes:

• Expose an activity or audit log that records system actions in machine-readable format, providing agents with a visible activity feed and corresponding JSON event log to track state changes.
• Add source references or input citations to generated outputs, allowing agents to trace data provenance and verify information origins.
• Pre-fill form inputs with sensible defaults wherever reasonable values exist, reducing the cognitive load on agents navigating configuration workflows.
• Clearly distinguish publicly accessible content from authentication-required areas, enabling agents to plan interaction paths without encountering unexpected authorization barriers.
• Gate destructive actions behind explicit, distinctly-labeled confirmation steps rather than same-styled buttons, giving agents clear signal before irreversible operations.
• Provide one-line summaries with expandable drill-downs for important results, allowing agents to efficiently scan information hierarchies and selectively retrieve detail.