Skip to content

Glossary

Anthropic’s AI coding CLI. SCC CLI runs the official Claude Code inside Docker sandboxes with team-managed policies and profiles.

A Claude Code work session running inside a Docker container. Sessions persist across restarts and can be resumed with scc start --resume.

A CLI that runs AI coding sessions inside an isolated environment (such as Docker) with controlled file, network, and policy access. SCC is a sandboxed coding CLI for Claude Code.

The Docker container that isolates Claude Code from the host system. Sandboxes provide security through containerization while allowing workspace access.

A configuration set for a specific team. Profiles define plugins, MCP servers, session settings, and delegation rules.

A git worktree created for parallel development. Each worktree has its own working directory and can run independent Claude sessions.

The directory mounted into the sandbox. Typically a git repository root where Claude Code operates.

The top-level configuration file hosted by an organization. Defines security policies, marketplaces, defaults, and team profiles.

Configuration specific to a team. Can be inline (in org config) or federated (external repository).

Optional per-repository configuration. Allows projects to add plugins and MCP servers within team-allowed boundaries.

Teams that store their configuration in external repositories, enabling independent management while respecting organization security policies.

The location of a federated team’s configuration (GitHub, GitLab, HTTPS URL).

Glob patterns that absolutely prevent usage of certain plugins, MCP servers, or Docker images. Cannot be overridden.

The system controlling which teams can add resources and whether projects can extend team configurations.

A time-bounded override for governance controls. Local overrides bypass delegation; policy exceptions bypass security blocks.

Permissions given to federated teams, controlling marketplace inheritance and additional marketplace definitions.

The scc-safety-net plugin that blocks destructive git commands like push --force and reset --hard.

An extension that adds capabilities to Claude Code. Installed from marketplaces.

A repository containing plugins. Can be official, organization-internal, or team-specific.

External services that provide context or tools to Claude. Types: HTTP, SSE (Server-Sent Events), or stdio (local process).

An MCP server running as a local process. Disabled by default due to elevated privileges.

Preview mode that shows what would happen without actually executing. Used with scc start --dry-run.

Command that shows effective configuration with source tracking. Run with scc config explain.

Inspection of installed plugins and their status. Run with scc audit plugins.

Cleanup of stopped containers or stale worktree entries.

Duration before an exception expires. Specified as 8h, 30m, 1d, etc.

Wildcard pattern using * for matching. For example, *experimental* matches any string containing “experimental”.

How long SCC caches remote configurations before re-fetching. Default: 24 hours.

Numeric code returned by commands indicating success or failure type. See CLI Overview.

An isolated environment for running AI coding assistants safely. SCC provides an AI code sandbox using Docker containers, preventing AI from accessing files or systems outside the designated workspace.

Policy controls that prevent Claude Code from executing dangerous commands or accessing restricted resources. SCC implements guardrails through Docker isolation, Safety Net (git protection), plugin governance, and network policies. See AI Coding Guardrails.

A container-based isolation environment for secure code execution. SCC uses Docker sandboxes to run Claude Code in an isolated environment where the AI can only access mounted directories and approved resources.

Practices and tools for safe AI-assisted development. Includes isolation (sandboxing), access control (plugin governance), history protection (Safety Net), and configuration management (team profiles). SCC provides comprehensive AI coding security for Claude Code.

Separating AI coding assistant execution from the host system to limit potential damage. SCC achieves isolation through Docker containers, controlled mounts, and policy enforcement.