MCP Fundamentals
How the Model Context Protocol turns APIs into tools an agent can call: connect the APIVerve MCP server to a real client, read tool schemas the way a model does, and lock down what an agent's key can spend.
MCP Fundamentals badgeComplete all 4 lessons to earn it. About 1.9 hours.
0 of 4 lessonsStart course
What you'll be able to do
- Explain tools, resources and prompts, and what a client and a server each own
- Connect the APIVerve MCP server to an editor or desktop client and prove it works
- Read input and output schemas, structuredContent and null or premium fields correctly
- Give each agent its own scoped sub-key with a bounded spend
- Treat tool results as untrusted input and keep prompt injection away from your actions
This course is one of three skill areas on exam APV-220, APIVerve Certified Developer: AI Agents: 10 of its 30 questions. See the full path
Syllabus
- What MCP is, and what a tool call actually returnsStart hereWhat the Model Context Protocol is, how tools, resources and prompts differ, and what a real MCP tool call sends and returns, shown with live API calls.
- Connect the APIVerve MCP server to your clientConnect the APIVerve MCP server to Cursor, VS Code, Windsurf, Claude Code or a stdio client, authenticate with OAuth or a key, and prove it works.
- Tool schemas and structured outputHow MCP input schemas, output schemas and structuredContent shape agent behaviour, why field descriptions matter, and how to read null and premium fields.
- Keys, scopes and safety for agentsGive each AI agent its own scoped API key, cap what it can spend, treat MCP tool results as untrusted input, and decide what an agent must never do.