LangSmith adds coding agent reporting to track spend by pull request outcome

·2026.11.19 00:00

Key point

The new feature attributes token costs to specific GitHub or GitLab pull requests, allowing teams to calculate the cost per merged PR and identify unattributed spend.

1 / 4

Details

  • LangSmith introduces coding agent reporting that ties agent spend to pull requests, enabling teams to measure ROI by connecting token costs to specific code outcomes.
  • A new dashboard aggregates costs across all traced agents, breaking down spend by PR state (open, closed, merged) and work type (features, bug fixes, investigation).
  • The platform supports tracing for Claude Code, OpenAI Codex, Cursor, GitHub Copilot in VS Code, and Deep Agents, consolidating data from multiple vendors into a single view.
  • Teams can use the metrics to coach engineers on thread management, evaluate model changes by comparing cost per merged PR, and align engineering and finance on cost visibility.
  • LangSmith for Coding Agents is in public beta on LangSmith Cloud, with setup involving installing agent-specific plugins and connecting GitHub.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.