claw0 is a progressive learning repository that teaches how to build an AI agent gateway from scratch in ten runnable Python sections.
The repository addresses the gap between basic agent tutorials that demonstrate a single API call and production systems that handle routing, persistence, and concurrent requests. It takes a cumulative approach where each section introduces exactly one architectural concept while preserving all prior code, allowing learners to understand how components integrate. The sections progress from a foundational agent loop and tool use through sessions and channels, then add gateway routing, configurable intelligence via system prompts, scheduled heartbeat tasks, durable message delivery, resilience patterns, and finally concurrent request handling with named lanes.
This repository suits developers who want to understand agent gateway architecture by building it incrementally rather than reading production code directly. The ten sections span roughly seven thousand lines of Python, with each section ranging from about 175 to 1130 lines. The approach works best for those comfortable with Python who want to see how concepts like JSONL-based session persistence, channel abstraction, binding tables for routing, and FIFO queue concurrency fit together in a working system. The repository includes code and documentation in English, Chinese, and Japanese with identical logic across languages.
Development activity shows consistent, focused iteration on the learning material. The project maintains clear section dependencies that build systematically from foundation through advanced topics, with each section's code remaining runnable and self-contained. Documentation is tightly integrated with code rather than separated, and the repository includes detailed architecture diagrams and section breakdowns that help learners understand both what they are building and why each component matters.
