DEV Community

Ed Mills
Ed Mills

Posted on • Originally published at indiestack.fly.dev

I asked Claude to build analytics from scratch. It cost 47,000 tokens. The alternative cost 2 lines.

#ai #productivity #opensource #webdev

I asked Claude to build analytics from scratch. It cost 47,000 tokens. The alternative cost 2 lines.

A real comparison: 47,000 tokens and three hours of back-and-forth versus 700 tokens and four minutes.

The Waste

I was building a SaaS dashboard and asked Claude to add analytics. What followed was an expensive lesson:

  • First prompt: Database schema with SQLite DDL for events, sessions, and page views (~15,000 tokens)
  • Second prompt: Event tracking middleware and utility functions (~18,000 tokens)
  • Third prompt: Dashboard with charts, graphs, and cohort analysis. Required follow-up fixes.

I had reinvented a wheel that hundreds of people had already reinvented, and done it worse than all of them. Final tally: roughly 47,000 tokens across three hours, producing code that wasn't production-ready without major additions like data retention policies and GDPR compliance.

The Discovery

I found IndieStack - a directory of indie tools with an MCP server integration. Setup was two commands:

pip install indiestack
claude mcp add indiestack
Enter fullscreen mode Exit fullscreen mode

This gave Claude access to a curated index of tools before it started generating code from scratch.

The Result

Next project, when I asked for analytics, Claude checked IndieStack first and suggested Simple Analytics - privacy-first, cookie-free, single script tag integration.

Token usage: ~700 tokens. Four minutes.

The Math

47,000 tokens vs 700 tokens. Thats the difference between building custom infrastructure and using an existing solution.

  • Building analytics from scratch: 40,000-60,000 tokens typically
  • Using an existing indie tool: ~700 tokens

And the custom build needs ongoing maintenance. The integrated solution doesn't.

The Broader Point

Auth. Payments. Transactional email. PDF invoices. Uptime monitoring. Background job queues. Every feature category has mature indie tools available. The problem is AI assistants don't have a way to discover them - they default to building from scratch.

The fix: give it the tool. Let it search first.

Installation takes thirty seconds. Next time you ask for a feature, your AI checks existing tools before writing boilerplate.

I wish I had had it three hours earlier.

Top comments (0)