Claude Code

CLAUDE.md and scoped rules provide first-party persistent instructions. They guide model behavior; they are not an enforcement boundary.

The reviewed project-instruction source does not define a complete cloud-surface loading contract.

Claude Code documents command, prompt, agent, HTTP, and MCP-tool hooks across lifecycle events and settings scopes.

The reviewed hook reference is insufficient to assert the same lifecycle contract on Claude Code's hosted surface.

Claude Code implements Agent Skills with progressive disclosure, manual or model invocation, scripts, resources, and product-specific controls.

The reviewed skills reference is insufficient to classify hosted discovery and invocation as equivalent to local clients.

Claude Code supports built-in and custom subagents with isolated context, tool and model controls, and nested delegation.

The reviewed subagent reference does not establish a complete hosted-surface contract equivalent to local clients.

Agent teams provide independent teammate sessions, a shared task list, mailboxes, and peer messaging, but Anthropic still labels the feature experimental.

The reviewed experimental agent-teams reference does not establish equivalent IDE and hosted coordination behavior.

Claude Code provides first-party MCP configuration and connection management for external tools and data sources.

The reviewed MCP source does not define a complete hosted capability and authentication matrix.

Claude Code provides built-in commands and invocable custom workflows; skills are now the recommended packaging model for multi-step commands.

The reviewed command reference is terminal-centered and does not establish identical IDE or hosted command availability.

Claude Code documents layered settings, permissions, managed policy, environment controls, and configuration precedence.

The reviewed settings source is insufficient to classify the hosted configuration and managed-policy surface as equivalent.

Claude Code combines human-authored CLAUDE.md context with machine-authored, repository-scoped local auto memory.

Claude's documented local auto-memory store is machine-local; the reviewed source does not establish an equivalent hosted store.