Create a multi-agent analysis pipeline in Claude Code
A data analyst had Claude Code assemble a five-part analysis system from a detailed specification.
- What they did
- Bronson, a data analyst, first had an AI app-idea agent generate a 713-line spec. He handed that document to Claude Code, which spent about five hours building the workflow: one agent for cleaning, one for analysis, one for charts, one for the written report, and one for the final summary, all coordinated by a @full-analysis controller that moved files between stages. Claude Code also produced tests and corrected its own errors, while Bronson mostly granted permission prompts.
- How it went
- The finished build reported 627 passing tests and an estimated $35 in API costs. It has not been tried on real production data yet, and an analyst still needs to check the outputs.
- Tips
- The hardest part was writing the specification, so give that plenty of attention. A Pro plan was enough, though Max was more comfortable because usage caps could slow the work.
Try it yourself with Claude Code
For [dataset file], set up a Claude Code project with five connected agents: one to clean the data, one to inspect it for patterns, one to make [number] charts, one to draft a full report, and one to condense everything into a one-page executive summary. Use a coordinator named @full-analysis to pass the files from one stage to the next, keep the source data unchanged, and save each stage’s output in its own folder. Before any action that would delete, overwrite, submit, send, or pay for anything, stop and ask me to approve it.
Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.
Get 3 agent uses like this every morning
With the day's top AI news. Free, and one click to unsubscribe.
Related agent uses
Have Claude Code build a cited stock research memo
A developer used Claude Code to gather filings and transcripts, then turn them into a source-backed memo and bear case on a holding.
Search large public archives with plain language
Claude Code can translate everyday questions into SQL or vector queries for Scry's read-only data index.
Turn new arXiv papers into nightly notes
A daily digest and nightly Claude Code summaries helped one person keep a small, organized notes folder for selected papers.
Have a coding agent check papers before each experiment
A Claude Code agent with paper search found useful ideas that a plain run missed while tuning a small language model.
Use Reddit threads as pre-code research
They used Claude Code with a Reddit MCP server to turn thread scanning into a fast research step before coding.
Turn a stack of articles into a cited summary
Have an agent compare many articles, pull out shared themes, and draft a source-cited summary document.