I built a company of bots
Most agent demos are one clever prompt. I wanted to see what happens when you organize agents the way a real company is organized: departments, specialists, and someone at the top deciding who does what. That is Botropolis: 20 specialist agents across 10 departments. Research has Scout. Code has Coder and Reviewer. Finance, legal, health, data, marketing, ops, security, support all have their…
Imagine a company where every department is represented by its own AI agent. This is the vision behind Botropolis, a company built by one individual. Each department is staffed by a specialist agent, such as Scout for research, Coder for coding, Reviewer for editing, and more.
The CEO agent serves as the central coordinator. It receives your request, decides which department is best suited, manages the work, and compiles a final report. Every agent is defined by a YAML file, specifying its name, title, specialty, model, tools, system prompt, and example tasks.
The system is designed to be transparent and accessible. A web-based user interface allows you to interact with the CEO, communicate with individual agents, and track analytics on each agent's performance. The CEO agent itself is a small, lightweight model trained from scratch on 300 synthetic questions and answers.
What sets Botropolis apart is not just the number of agents, but the system it represents. Routing is crucial - sending the right request to the right agent. The YAML specs make the agents easily reviewable and understandable. Performance analytics reveal insights that help in optimizing the system, just like a manager would watch budgets and performance metrics.
The creator of Botropolis learned that organization and management are key. Even with 20 agents, they are ineffective without proper routing. YAML specs improve transparency and reviewability. Analytics can shift how you build and manage these systems. Future plans include streaming responses, integrating tool execution directly into the model clients, and fine-tuning more agents. The entire project is open source under the MIT license. You can explore the code and contribute at https://github.com/kagithamanoj/botropolis.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
