How we use Insights to improve the agent behind Insights

Datadog’s Agent Observability Insights analyzes AI agent traces to identify recurring cost, reliability, and tool-use issues, explain their causes, estimate their impact, and recommend evidence-based improvements. Examples from Datadog’s own agent included overly broad searches that returned tens of thousands of unnecessary tokens, prompt structures that reduced provider caching by placing task-specific content before stable instructions and tool definitions, repeated invalid tool calls missing required arguments, and requests for resource names not recognized by tools. Recommended fixes included narrowing searches, using compact previews that retain retrieval identifiers, restructuring prompts around stable shared prefixes, validating resource names, and preventing retries that cannot succeed without new information. The platform can also send findings and trace references to coding agents through its “Fix with Bits” workflow, enabling them to prepare tested pull requests, such as one that introduced compact search previews. Insights is generally available with a free tier covering up to 40,000 LLM spans monthly, allowing teams to assess fixes through replayed runs and monitor whether problems recur.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.