MCP server integration
Integrate MCP servers into Claude Code and agent workflows.
This subtopic (2.4) sits in Tool Design & MCP Integration (D2) on Anthropic's Claude Certified Architect — Foundations (CCAR-F) exam. The bank holds 19 practice questions here — 5 easy, 9 medium, and 5 hard — with 6 free to try, answers and explanations included. 3 of the free questions are below; the rest are in the practice stream.
What the exam tests here
- MCP server scoping — project-level .mcp.json vs user-level ~/.claude.json
- community MCP servers vs custom implementations for standard integrations
- MCP resources as content catalogs to reduce exploratory tool calls
- environment variable expansion in .mcp.json — credential management
Free practice questions: MCP server integration
Your team wants every developer to automatically have access to the get_customer and lookup_order MCP tools when working on the support agent project. Which configuration approach achieves this?
Show answer & explanation
Correct answer: A. Add the tools to .mcp.json in the project root and commit it to version control
Project-scoped MCP servers are stored in .mcp.json at the project root and shared via version control, making them automatically available to all team members working in that project. Local scope (~/.claude.json) is per-machine and not shared. Environment variables and local scope cannot distribute tool access across the team.
Your support agent's get_customer and lookup_order MCP servers must be available to you on every project you open in Claude Code on your laptop, including private experiments outside the support repo. Other teammates working in the support repo should not see these servers when they clone — your laptop holds personal API keys to a sandbox tenant. Where should these MCP server definitions live?
Show answer & explanation
Correct answer: A. In ~/.claude.json on your laptop with user scope, so they load for any project you open without affecting teammates.
User-scoped MCP servers in ~/.claude.json load across every project for that user without committing anything to a repo, matching the 'all my projects, no teammates' constraint. 'In a committed .mcp.json' is project-shared, which the question explicitly rules out. '.claude/settings.local.json' holds local settings overrides, not MCP server definitions. 'claude mcp add with --scope project' writes to the project file, which would expose the servers to teammates.
Your customer support agent calls lookup_order repeatedly to discover what order fields exist before forming queries. Which MCP capability best eliminates this exploratory overhead without changing the tool interface?
Show answer & explanation
Correct answer: B. Expose an MCP resource listing the order schema so the agent reads field definitions once before invoking lookup_order
MCP resources serve as application-driven content catalogs. Exposing the order schema as a resource lets the agent read field definitions once via resources/read, eliminating repeated exploratory lookup_order calls. The other options add tools, adjust timeouts, or cache responses — none exploit the resource catalog pattern the spec defines.
3 more free questions on this subtopic in the practice stream, plus 13 in the full bank. Keep practicing →