Agentic AI Interview Questions
Dashboard
0%
1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
51 small wins to finish your pathNext question →

Q5EasyConcept

What are the core components of an agent?

30-second answerSay your answer out loud first, then reveal.
  1. Model. The LLM that reasons and decides. Choice affects tool-use reliability, latency and cost. Many systems mix models, e.g. a strong model for planning and a cheap one for simple sub-tasks.
  2. Instructions. The system prompt defines role, goal, constraints, tone, when to ask for help, and when to stop.
  3. Tools. Functions the agent can call: retrieval, APIs, code execution, browser, other agents.
  4. Memory.
    • Short-term: the current conversation and scratchpad (the context window).
    • Long-term: facts, user preferences and past episodes stored externally (vector DB, key-value store) and retrieved when relevant.
  5. Planning / orchestration. The loop logic: ReAct, plan-and-execute, graph-based state machines, max steps, retries.
  6. Guardrails. Input/output filtering, tool permission checks, human approval for risky actions, PII handling.
  7. Production layer. State checkpointing (resume after crash), tracing (see every step), evaluation, cost tracking.
Core components of an agent: instructions, guardrails and tracing feed the agent, whose LLM reasoning loop exchanges data with short-term memory, tools and long-term memory.

Common mistakes

  • Listing only "LLM + tools + memory" and forgetting guardrails and observability, which is what separates a demo from production.

Follow-ups to expect

  • Which component fails most often in production? A good answer is tools and context, not the model.
PreviousWorkflow vs agent: when would you NOT use an agent?

Every expert started right here.