claude-crew

Claude Crew provides a collection of ready-to-use AI agent personas for Claude Code. Each agent is a specialized expert with a defined personality, methodology, and output format, designed to be dropped into your Claude Code workflow. Use these agents to power tasks such as code reviews, test generation, documentation, refactoring planning, API design, DevOps guidance, and more, by pointing Claude Code at the agents directory and invoking the appropriate persona for the job. To start, clone the repository and place the desired agents into your Claude Code environment; Claude will automatically detect and apply any agent definitions found under .claude/agents/ or the corresponding fun and swarm templates. You can selectively pull individual agents (for example, a code-reviewer) into your project and leverage their specialized approaches directly in your coding and review processes.

Prerequisites:

• Git
• Node.js installed locally if you plan to use the MCP route via npx (optional depending on your setup)

Installation steps:

1. Clone the repository: git clone https://github.com/hlsitechio/claude-crew.git
2. Navigate into the project (optional for exploration): cd claude-crew
3. Use an MCP-compatible runner (example with npx):
   • Ensure you have npm installed and Node.js available.
   • Install or run the MCP server wrapper as described by your MCP tooling using the provided package name if applicable. For Claude Code integration, you typically point Claude Code at the agents directory or copy agents into your project’s .claude/agents/ path as shown in the Quick Start.
4. If you want to experiment with a subset of agents, copy the desired agents into your project’s .claude/agents/ directory (or the equivalent path you use in Claude Code).
5. Launch Claude Code and verify that the agents are detected and available for selection in your tasks or workflows.

Tips and caveats:

• Agents live as markdown definitions within the repository (under agents/ and fun/ directories). Claude Code reads these definitions to adopt persona, methodology, and output format.
• You can mix and match agents (e.g., Code Reviewer + Test Writer + Doc Generator) within a workflow or swarm to achieve multi-faceted analysis.
• For best results, ensure the agents you deploy have clear output formats and rules to minimize ambiguity in Claude Code responses.
• If Claude Code does not automatically detect agents, confirm the presence of a .claude/agents/ directory in your project and that your version of Claude Code supports custom agents via markdown definitions.