Airsc is free for everyone. If it helps you, support it with $2–$20.

Support

Browse the library

Filters1
AgentSafe

house-agents

houseworthe

A collection of specialized Claude Code sub-agents designed to manage context efficiency by delegating heavy operations and returning condensed results to the main conversation.

42MIT
AgentCaution

claude-code-subagents

mylee04

This collection provides 31 gamified, specialized AI sub-agents for Claude Code that work as a development team and level up.

33MIT
AgentSafe

agency-agents

Raheel2774

This is a collection of 200+ specialized AI agent personas for various AI coding assistants, easily installable via a native app.

33MIT
AgentCaution

Claude-Code-Subagents-Collection

gensecaihq

Offers a collection of over 75 specialized Claude Code sub-agents designed for enterprise-grade software development tasks, enhancing accuracy and efficiency.

30MIT
AgentSafe

agent-dispatch

ginkida

This MCP server and CLI allows Claude Code agents to delegate tasks to other agents across project directories for multi-agent orchestration.

30MIT
AgentCaution

kiro-agents

Theadd

Installs an agent system into Kiro IDE for creating custom agents, switching between vibe and spec modes, and keeping persistent memory across sessions.

30
AgentCaution

MCP-Developer-SubAgent

gensecaihq

A specialized framework for Model Context Protocol (MCP) development, offering 8 Claude Code sub-agents, security hooks, and production-ready FastMCP server templates.

28MIT
AgentCaution

cli

prompt-scheduler

This CLI tool automates the scheduling and execution of prompts for AI agents like Claude Code, intelligently detecting and waiting for usage limit resets.

23MIT
AgentSafe

ai-bwrap

didvc

This tool runs AI coding agents like Claude Code inside a bubblewrap sandbox, limiting their access to only the current working directory for enhanced security.

18MIT
AgentCaution

claude-code-jsonl-editor

didvc

An interactive editor for Claude Code's conversation files, enabling efficient prompt engineering through direct modification of AI responses.

14MIT