Tool Calling Interview Questions

Medium AI Agents interview question on tool calling design best practices — architecture, trade-offs, eval, and production patterns.

AI Agents are the hottest topic in 2025–2026 GenAI interviews, but most candidates confuse agents with chatbots. Interviewers at OpenAI and Anthropic want you to articulate the agent loop — perceive, plan, act, reflect — and explain when tool use justifies agent complexity over a simple chain.

The Model Context Protocol is Anthropic's open standard reshaping how LLMs connect to tools and data. If you're interviewing for Claude ecosystem roles or AI platform engineering, expect MCP questions. Understand host vs server architecture and how MCP differs from ad-hoc function calling.

A production AI coding agent typically includes:

This is an enterprise platform design question spanning agents, MCP, governance, and observability.

MCP (Model Context Protocol) defines a clean separation:

MCP multiplies agent power and attack surface.

Prompt injection is the #1 security interview topic for agents that read untrusted text.

Medium Claude cost question on prompt caching breakpoints and hit rates.

Medium agents platform question on reusable skill packs and governance.

Hard agents ops question on timeouts, circuit breakers, and budget protection.

**Mechanism** Server caches designated prompt prefixes; subsequent requests reuse if prefix identical.

**Mechanics** Assistant message may include tool_use blocks with id, name, input JSON; user message returns tool_result tied to id.

**Session** Persistent realtime connection with session.update for instructions/tools; handle reconnect.

**Concept** Responses treat interaction as structured items (messages, tool calls, outputs) with native support for reasoning models and built-in tools — reducing glue code.

**Unit layer** Pure functions: conditional edges, state merge logic, parsers — no graph compile needed.

**Retry patterns** Tenacity-style retries inside node or dedicated retry wrapper node incrementing state.retry_count. Exponential backoff for transient MCP errors.

**Cycle design** Classic ReAct loop: agent → tools → agent. LangGraph makes cycle explicit vs hidden while True loops. Conditional edge from tools checks whether to continue.

Company-based MCP questions assess platform leadership, not wire protocols alone.

Dangerous tools are why agents aren't auto-root on production.

Rate limiting for agents differs from human API usage — bursts are larger and less predictable.

Migration questions test program management as much as protocol knowledge.

Long-running tools expose gaps between synchronous tool calling and real workloads.

Database MCP is high risk/high value — interviewers probe your safety instincts hard.