Claude Code just got more introspective with the /insights command, a diagnostic tool that generates detailed reports on how developers are actually using AI assistance in their day-to-day work.

What Gets Tracked The command surfaces four core metrics: which projects consumed your attention (with percentage breakdowns), what tools you relied on most heavily—Edit, Bash, and Read show up frequently based on the example output—and critically, where friction crept into your workflow. The "Common friction" section flags recurring pain points like test flakiness or repeated permission prompts that slow down development cycles.

Reading Your Own Usage Patterns The example output shows a 30-day window with 142 sessions logged across three primary projects: api (61%), web (24%), and infra (9%). This kind of granular visibility lets developers answer questions they couldn't previously quantify—like whether they're actually delegating repetitive tasks to Claude or still doing too much manually. The data also highlights tool preference patterns that might reveal inefficient workflows worth restructuring.

The Quarterly Review Angle Anthropic positions /insights as particularly useful for end-of-quarter self-assessments. Instead of guessing where your time went during AI-assisted coding sessions, you get objective metrics to reference when writing performance reviews or justifying tool investments to leadership. This transforms Claude Code from a purely reactive assistant into something that also provides retrospective analytics on your development habits.

Why This Matters for the Ecosystem We're watching Anthropic build out Claude Code as more than just a CLI wrapper around an LLM—it's becoming a full developer workflow platform with telemetry baked in. Usage analytics like this create accountability and optimization opportunities that pure chat interfaces lack. Developers can iterate on their own practices using data rather than gut feelings, while teams might eventually aggregate these insights across organizations to establish baseline efficiency benchmarks for AI-augmented development.

Key Takeaways

  • /insights tracks project distribution, tool usage, and recurring friction points over configurable time windows
  • The command targets individual developers seeking objective self-assessment data for reviews or process optimization
  • Anthropic appears to be positioning Claude Code as a platform with analytics capabilities, not just an AI pair programmer

The Bottom Line This is quiet but significant—Anthropic is building observability into the developer experience itself. When your CLI starts generating usage reports, you've crossed from tool into platform territory. Whether you're ready for that or not, the data doesn't lie about where your time actually goes.