claude-code-configlisted

# Configure Claude Code Configuring Claude Code means shaping how the harness itself behaves through its JSON settings files and the surfaces they govern: what the agent is **allowed to do** (permissions), what runs **automatically** on harness events (hooks), the **environment** every session sees (env), and which **external tool servers** are wired in (MCP). Whether you are standing up a configuration from scratch or changing one existing setting, the job is the same shape: put the *right change* in the *right file* at the *narrowest scope that works*, keep the JSON valid, and verify it actually took effect. Answer from this skill's reference content, not from a half-remembered key name. Inventing a settings key that does not exist produces a file the harness silently ignores. ## Which surface is this? Every configuration request lands on one of these surfaces. Find the row, read the reference. | The request is to… | Surface | Read | |---|---|---| | Allow/deny/ask a tool or command; move a rule between scopes; tighten an over-broad rule | Permissions | `references/permissions.md` | | Run something **automatically** whenever an event happens ("always", "every time", "whenever X") | Hooks | `references/hooks.md` | | Add an environment variable, or a change isn't taking effect | Env / verification | `references/settings-files.md` + `references/verification.md` | | Register / enable / disable an MCP server | MCP | `references/settings-files.md` (MCP section) | | Change an