agent-core
The endpoint-agnostic half of an OpenAI-compatible agent loop.
Tool-schema compatibility across endpoints, on-demand tool loading, one-shot side tasks, run events, and a pooled client — the parts of an agent loop that have nothing to do with which model is answering.
It was extracted from three servers that had each written it separately — kanban-server, task-server and min-agent — after the copies drifted far enough that a fix in one was a bug still live in another.
Stack: TypeScript. Published as @cubicecho/agent-core.
Full documentation lives in the project README: https://github.com/cubicecho/agent-core