search-first

To find existing libraries before writing custom code, systematize your research workflow by analyzing needs, searching parallel resources via agents across npm, PyPI, and GitHub, evaluating candidates, and deciding whether to adopt, extend, or build.

Evaluating dependency maintenance and licensing involves analyzing candidates found through parallel agentic searches across npm, PyPI, and internal MCP servers, measuring functionality and maintenance health to inform an adopt, extend, or build custom decision.

Yes, you can search npm, PyPI, and GitHub in parallel when adding a new API client by invoking a dedicated researcher agent that comprehensively scans these repositories for existing, well-maintained solutions.

Yes, the research workflow integrates with planner and architect agents, supporting informed decision-making by passing evaluated candidates and need analysis directly into the downstream architectural planning process.

You should build a custom solution instead of using existing code patterns when comprehensive parallel searches across npm, PyPI, and GitHub yield no candidates meeting your required functionality, maintenance, and licensing criteria.

To systematize code discovery and prevent redundant effort, follow a structured workflow: analyze needs, invoke parallel agentic searches across repositories, evaluate options based on functionality and licensing, and formalize an adopt, extend, or build decision.