Your coding agent bill doubled. Here’s how to fix it.

LangChain says LangSmith now provides a unified way to trace coding-agent sessions from Claude Code, Codex, Cursor, GitHub Copilot Chat, Pi and OpenCode, with normalized data on token usage, costs, tool calls, subagents, errors and timing. The company argues that spending becomes difficult to manage when teams use multiple agents whose native dashboards record activity differently, making it hard to compare the cost and value of a feature or workflow. LangChain presents a four-part approach: centralize spend visibility, standardize cost measurements, use its Engine to identify waste such as redundant tool calls, and apply LLM Gateway caps and routing controls. The offering is primarily positioned for teams using more than one coding agent; LangChain acknowledges that a single tool’s native dashboard may be sufficient for teams that have standardized on it. The post cites examples of sharply increasing enterprise AI costs, but does not provide independent evidence or further detail for those claims.