Free — Already in Claude Code
Your Coding DNA in One Command
Claude Code Insights analyzes how you actually use Claude Code and tells you what you're great at, where you lose time, and exactly how to improve.
Open your terminal. Type this.
claude insights
That's it. 1-2 minutes. No setup, no API key, no installation. Already built into Claude Code.
It tells you two things
🟢
The Good Stuff
What you're doing that most people aren't. Your strengths, your edge, the workflows where you're ahead.
🔴
The Bad Stuff
Where you lose time, what breaks, which friction patterns cost you hours. And specific fixes for each one.
⚡
Quick Wins
Copy-paste CLAUDE.md rules generated from YOUR usage. Instant improvement.
🚀
What's Next
Advanced workflows matched to your skill level. The 10x moves you're ready for.
Real Examples From My Report
Here's what the report actually found in my usage (1,100+ messages across 124 sessions):
My actual Insights report — the "At a Glance" section
🟢 What's working
"You've built an impressive operational rhythm — running full cycles from database queries through multi-file fixes to production deployment in single sessions. Your parallel agent coordination pattern for batch bug fixes is genuinely advanced usage, and your habit of creating reusable skills means each session compounds on the last."
The report recognized my workflow isn't just "ask Claude to fix things." It's a system — parallel agents, reusable skills, compounding sessions. That's what separates power users from everyone else.
🔴 What's hindering you
"Wrong approaches are your biggest time sink — Claude frequently picks the wrong implementation path or oversteps scope (like building prototypes when you asked for brainstorming only), forcing you to intervene and redirect."
80 "wrong approach" events in one month. Hours lost to Claude building when I said "just think." The fix: I added a design preview gate — Claude shows me the plan BEFORE building. One rule, hours saved every week.
🔴 External walls I kept hitting
"Sessions sometimes hit avoidable walls with external services (GoDaddy API limits, Google Cloud permissions, broken shell state from prior sessions) that could be pre-validated before diving into implementation."
I'd spend 30 minutes building, only to discover GoDaddy's API requires a paid plan or Google Cloud permissions expired. The fix: pre-validation checks at session start. Verify API access BEFORE writing code.
⚡ Quick wins to try
"Try using hooks to auto-run your /status check and TypeScript build validation at session start and after edits. Also consider creating a custom skill that encodes your parallel-agent bug-fix coordination pattern as a reusable /command."
Two concrete suggestions tailored to MY patterns. I implemented both — /status now auto-runs, and /tdd-fix is a reusable skill. The report told me exactly what to build.
🚀 Ambitious workflows — where you're heading
"Your parallel agent pattern can evolve into a proper coordinator system — a lead agent that assigns scoped tasks, prevents file conflicts, merges results, and runs a unified build before surfacing anything to you."
The report predicted Agent Teams before I even used them. Now I run 3 parallel agents building features on separate branches, coordinated by a team lead. The report showed me where I was heading before I got there.
What To Do With Your Report
1
Read the "Where Things Go Wrong" section
This is your biggest time sink. The report doesn't just show the problem — it suggests the specific fix. A CLAUDE.md rule, a hook, a skill. Copy it.
2
Copy the suggested rules into your CLAUDE.md
The report generates rules based on YOUR friction. Paste them into your project's CLAUDE.md file. Claude Code reads this on every session — your experience immediately improves.
3
Try one thing from "What's Next"
The report identifies advanced workflows matched to your level. Agent teams, custom skills, parallel execution patterns. Pick one and try it this week.
Pro tip: Run it again in 2-4 weeks. Your friction patterns shift as you improve. The report evolves with you.
Need ~20+ sessions for a meaningful report. If you just started with Claude Code, keep building for a few weeks and come back.
Level Up: Discuss Your Report With Claude
The report gives you data. Claude helps you act on it. Open Claude.ai or Claude Code and paste your report with one of these prompts:
💬
"What should I fix first?"
Paste your report and ask: "Based on this Insights report, what are the top 3 things I should change in my workflow? Give me the specific CLAUDE.md rules to add." Claude reads your friction patterns and writes the rules for you. Copy-paste into your config.
💬
"Build me a skill for my top friction"
Ask: "My biggest friction is [X]. Build me a Claude Code skill that prevents this from happening again." Claude creates a reusable /command that encodes the fix as a permanent workflow — not just a one-time rule.
💬
"Compare me to best practices"
Ask: "Here's my Insights report. Compare my workflow to the top 1% of Claude Code users. What am I missing? What should I try next?" Claude maps your patterns against advanced techniques — agent teams, hooks, custom skills, scheduled tasks — and tells you what to adopt.
💬
"Create my improvement plan"
Ask: "Based on this report, create a 2-week improvement plan. Week 1: fix my top friction. Week 2: try the ambitious workflows. Give me specific daily actions." Claude turns the report into an actionable roadmap — not just insights, but a plan.
My story: I pasted my Insights report into Claude.ai and asked "how do I become more agentic?" That single conversation started something bigger — I'm now building a full AI agentic transformation of how I work. 22 lessons learned, 55+ custom skills, 7 projects running in parallel, quality gates that catch problems before I see them, and a system that gets smarter every session. The Insights report was the spark. I'm sharing everything I learn along the journey — follow along.
Did You Find This Guide Useful?
Share it with people who are learning AI (and specifically Claude Code). We'll be creating more guides like this.
