Skip to content

🧠 Overview

Kebogyro is a fast, async-first orchestration layer designed to make it easy to build LLM-powered applications with real tool-calling capabilities.

Inspired by:

  • 🛠 Gyro Gearloose — a symbol of mechanical cleverness
  • 🎶 Kebogiro — a Javanese ceremonial gamelan suite

Together: kebogyro combines structure, power, and orchestration — but for AI agents.

✨ Philosophy

  • Async-native: All I/O is async def, suitable for modern event loops.
  • Composable: Plug in your own LLM provider, cache backend, or tool logic.
  • OpenAI-compatible: Use OpenAI-style tool specs, but with any provider.
  • No framework lock-in: Integrate with anything — FastAPI, Flask, Celery, etc.

🧩 Architecture

  • LLMClientWrapper: Base abstraction over LLM APIs
  • SimpleTool: Describe tools with arguments and behavior
  • BBServerMCPClient: Protocol-aware adapter to remote tool bridges
  • create_agent: Glues it all together (LLM + tools + MCP)
  • AbstractLLMCache: Optional caching layer for tool specs/results

📦 When to Use

Kebogyro is perfect for:

  • Multi-provider LLM routing and orchestration
  • Tool-enabled agents in async web backends
  • LLM interfaces where caching or streaming is required
  • Building OpenAI-compatible tool calling into any environment

If you're building a serious LLM agent system that talks to code, services, or other users — Kebogyro is your async Swiss Army knife.

Next → Getting Started