shanraisshan/claude-code-multi-agent-orchestrartion

This project demonstrates how to coordinate multiple AI agents working together inside Claude Code (Anthropic's command-line coding assistant). The specific example it ships with fetches the current temperature for the capital city of every country in the world, all 195 of them, and then calculates a global average. You trigger the whole thing by typing a single command. When you run /orchestrate, the system spins up 195 individual "weather agents" in parallel. Each agent is responsible for looking up the temperature in one country's capital. Once all of those agents finish their lookups, a separate "writer agent" collects all the results and saves them to a markdown file. Finally, an "average agent" reads that file, computes the mean temperature across every country, and writes the result to a second file. The agents are color-coded (yellow for the country agents, red for the writer, green for the final calculator) so you can visually track which stage is running. The audience here is developers or tinkerers who already use Claude Code CLI and want to see a concrete example of multi-agent orchestration, meaning multiple AI tasks running side by side and then feeding their results into a pipeline. A practical use case: imagine replacing "weather" with "competitor pricing across 195 markets" or "sentiment analysis for 195 product reviews." The temperature demo is a stand-in for any problem where you need to gather data from many sources in parallel, aggregate it, then derive a summary. The README doesn't go into detail on how to install or configure the project, what dependencies are required, or how to customize the agents for your own use cases. It focuses entirely on showcasing the architecture and the output. The main tradeoff is that spinning up 195 parallel agents is resource-intensive, fine for a demo, but you'd want to think carefully about rate limits and cost before adapting this pattern to production-scale workloads.