Subagents each run in their own context window with their own tools — so they don't clog your main session and can run in parallel. Describe a task, pick a pattern, and get the ready-to-save .claude/agents files plus the orchestration prompt to drive them.
Free · no signup · runs in your browser
The .claude/agents files
Orchestration prompt · paste to your lead Claude
How to use it: save each block as .claude/agents/<name>.md in your repo (create the folder if needed). Claude Code auto-discovers them; the lead session delegates to them by their description, or you invoke one explicitly. Then paste the orchestration prompt to kick the whole thing off. Parallel subagents run in separate context windows, so a big fan-out won't blow your main session's context.
Want a vetted agent library?
Reviewer, tester, researcher and refactor agents tuned for real repos — in the power-user pack.