Building an Open Source Developer Experience Agents will Love

As open-source library developers increasingly interact with AI coding agents, the DBOS team recognizes the necessity of optimizing software and documentation for machine comprehension. They initially improved documentation accessibility by generating plain text files for AI use, but found that documentation alone was insufficient for timely updates. Attempts to create comprehensive prompts for agents resulted in context bloat, prompting a shift to a "skills" system that selectively provides information, mitigating overload and maintaining agent efficiency. Despite improvements, some challenges remain, such as agents struggling to choose the correct skill, particularly with vague user prompts. Additionally, for debugging, the team developed a Model Context Protocol (MCP) server, enabling agents to autonomously access application observability and telemetry information, thereby facilitating independent debugging and analysis of running systems. The DBOS team continues to refine these approaches to enhance the interaction between AI agents and their durable workflows library.