Run data cleaning to executive summary with agent pipeline
Built five Claude Code agents (cleaner, analyzer, visualizer, report writer, summarizer) chained by an orchestrator for end-to-end data analysis.
Done withClaude Code
- What they did
- Bronson, a data analyst, used an AI app-idea agent to produce a 713-line specification. He then pointed Claude Code at it. Over about five hours Claude Code built five agents plus a @full-analysis orchestrator that passes files between them. It wrote its own tests and fixed its own errors. He only approved permissions, about 20 times.
- How it went
- The build produced 627 passing tests and cost about $35 in estimated API usage. It has not yet been run on real work data, and outputs still need analyst review.
- Worth knowing
- The spec was the hard part. Pro plan is enough, though usage limits slow you down; he moved to Max for convenience.
Try it yourself with Claude Code
Set up a chain of subagents in this project for analyzing [dataset file]: a cleaner, an analyzer, a visualizer, a report writer, and a summarizer, run in order by an orchestrator. Have them clean the data, find the main patterns, make [number] charts, write a full report, and finish with a one-page executive summary. Save each stage's output in its own folder and keep the original data unchanged.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: bronsonelliott.substack.com · Oct 11, 2026 · Submitted by Bronson Elliott
Related use cases
Turn tickets into merged pull requests
Pulls tasks from GitHub Issues or Linear, then runs Claude Code or Codex in isolated Kubernetes pods to carry each one through to a PR.
Email Houston dealerships to shop for a car
Cowork built a dealership tracking sheet, sent outreach emails, filled web forms and managed replies; a separate 100+ receipt expense job hit limits and moved to Claude Code.
Research and draft cold outreach with Claude Code
Claude Code handles research and drafting for cold outreach while a separate platform sends at scale.
Run persistent financial research with an agent
LangAlpha is a Claude Code-style agent for investing research that keeps persistent workspaces and memory so analysis such as models and comps can be updated across sessions.
Custom agent
Run first-pass topic research, then verify references
A commenter uses Manus for first-pass research, then has Genspark double-check every reference and flag problems.
Produce cited research briefs with simulated peer review
Feynman multi-agent system searches papers and the web, synthesizes findings, runs simulated peer review, verifies citations, and outputs a cited brief.
Custom agent
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.