Making agentic token costs visible in production

In the realm of AI engineering, managing token costs has emerged as a significant challenge, as excessive token usage can inflate production costs and obscure productivity measures. Datadog's 2026 State of AI Engineering report highlights this issue, noting a substantial increase in token usage per request among its customers. The text explains that token costs accumulate across sessions, tool calls, and users, emphasizing the importance of visibility in tracking these costs to prevent unexpected expenses. It outlines strategies for reducing token costs, such as trimming tool catalogs, using prompt caching, and managing session history through windowing or summarization, while also addressing the need for governance through token budgets and the use of tools like Datadog Agent Observability to monitor and enforce cost controls. The importance of balancing efficiency with agent capability is underscored, as excessive pruning or compression might limit an agent's functionality, suggesting that these optimizations should be guided by data-driven assessments of their impact on performance and cost.

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.