Project #4 in a learn-MCP series: an agent that hires help. spawn_agent is runAgentLoop called recursively; every sub-agent's gated call bubbles to ONE human queue; and the eval suite says whether delegating was actually worth it.
3 files for Claude Code, Codex and OpenCode: learn-mcp-agent-crew, learn-mcp-agent-crew AGENTS.md, learn-mcp-agent-crew CLAUDE.md — 159 tokens loaded in every session.
AGENTS.md A 154 tok CLAUDE.md A 5 tok .claude/settings.json A — These files are ketankshukla/learn-mcp-agent-crew's own configuration — they tell Claude Code, Codex and OpenCode how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.