Skip to content

Category

Agentsrules, skills & MCP servers

Agents are autonomous operators you hand a goal and let run — subagents for Claude Code, custom modes for Cursor and Codex, or graph nodes for frameworks like LangGraph and CrewAI. A well-built agent has a sharp role, a tool allowlist, and a clear stopping condition; a badly built one loops, burns tokens, and edits files it shouldn't. That gap is exactly what the quality score below measures: we grade agents on how tightly their description scopes the work, whether they declare their tools, and how fresh and complete the definition is. Every agent is attributed to its real GitHub author, so before you let one touch your repo you can read who wrote it and what else they ship. Drop an agent definition into your project, and your assistant can dispatch it for a bounded task — a security audit, a refactor pass, a research sweep — without you babysitting each step. Browse the top agents below ranked by quality, or open the full catalog to filter by framework and the job each one is built to do.

Top agents right now

Ranked by quality score — freshness, schema completeness, and review signals. Refreshed daily.

Showing the top 24 of 47 agents.

Browse other categories