What is Claude Code?

2 min read Updated

Claude Code is Anthropic's agentic coding tool, available as a terminal CLI, IDE integration, and desktop app. It runs Claude in an agent loop that reads, edits, and executes code in a developer's environment, and it acts as an MCP client that can connect to external tools via Model Context Protocol servers.

WHY IT MATTERS

Claude Code matters to MCP governance because it is one of the most widely deployed MCP clients in engineering organisations. Developers attach MCP servers to it at three scopes: user-level configuration, a project-level .mcp.json checked into the repository, and enterprise-managed configuration. Each connected server contributes tools the agent can call autonomously during a session.

For fleet operators, the relevant surface is:

  • Configuration sprawl — per-developer and per-repo server lists are easy to add and hard to inventory, a common source of shadow MCP.
  • Permissions — Claude Code prompts for approval on sensitive actions by default, with allow/deny rules configurable in settings; teams tune how much runs unattended.
  • Enterprise controls — administrators can deploy managed-settings.json and a managed MCP configuration that pins or restricts which servers users may connect (allowlist/denylist, or managed-only mode), enforced ahead of user and project settings.
  • Delegation — sessions can fan out into subagents, multiplying concurrent tool callers under one user.

Comparable agentic coding tools (Cursor, Codex CLI, and others) raise the same questions; Claude Code is simply where many teams meet them first.

Running agents against MCP servers? Route them through PolicyLayer and every tool call is checked against policy first.

PUT POLICY ON YOUR TOOL CALLS →

Enforced before the call runs. Nothing to install.

HOW POLICYLAYER USES THIS

PolicyLayer governs Claude Code's MCP traffic when its servers are routed through the gateway: the developer's .mcp.json points at PolicyLayer endpoints with a per-person scoped token, and every tools/call Claude Code makes is evaluated against the team's policies before it reaches the upstream server. That gives platform teams one enforcement and audit point across Claude Code, Cursor, Codex, and any other MCP client in the fleet.

FREQUENTLY ASKED QUESTIONS

Is Claude Code an MCP client or an MCP server?
Primarily a client — it connects to MCP servers to gain tools. It can also be exposed as an MCP server itself, letting other applications drive Claude Code over the protocol.
Where does Claude Code store MCP server configuration?
At user scope in the developer's home directory, at project scope in a .mcp.json file in the repository, and at enterprise scope via managed configuration files deployed by administrators.
Can an organisation restrict which MCP servers Claude Code connects to?
Yes. Enterprise-managed settings support allowlists and denylists for MCP servers, and a managed-only mode that blocks servers not defined by administrators. A gateway adds enforcement on the calls themselves.

FURTHER READING

Let agents act without letting them run wild.

Route your MCP servers through PolicyLayer and every tool call is checked against your policy before it runs — allow, deny, or require approval. Per-identity grants. Full audit log. Live in minutes.

Free to start. No card required.

43,000+ MCP servers and 220,000+ tools scanned and risk-classified.

// GET IN TOUCH

Have a question or want to learn more? Send us a message.

Message sent.

We'll get back to you soon.