# Configuring and Extending Agents

This chapter covers making an agent fit your project and your research:

- {doc}`Configuring Agents for Your Project <configuring_agents>`: project instructions, skills, MCP servers, and subagents, plus the guardrails that hold whatever the model decides: permission rules, hooks, caps on unattended runs, and git as an undo layer.
- {doc}`Building Custom Tools and MCP Servers <building_custom_tools>`: packaging your own workflows as skills, and giving an agent your data and cluster tools through a small MCP server.