IICTI Certified AI Agent Developer

Holders demonstrate that they can apply and evaluate agent architectures and control-flow design, tool use, function calling and MCP integration, context engineering and agent memory and multi-agent orchestration and delegation, plus 7 further areas, in realistic professional scenarios, as defined in the AQSCERT scheme for AI Agent Developer.

Who it is for: Professionals working, or preparing to work, in artificial intelligence roles.

1. 01Agent architectures and control-flow design10% of the exam

   • Architect agent control flows that balance autonomy, determinism and cost for a given task profile.
   • Scope: ReAct, plan-and-execute, reflection loops, state machines and graphs (LangGraph), workflows versus autonomous agents, termination conditions and loop budgets.
2. 02Tool use, function calling and MCP integration10% of the exam

   • Design tool interfaces and MCP integrations that agents can use reliably and safely.
   • Scope: JSON-schema tool definitions, Model Context Protocol servers and clients, tool selection accuracy, idempotent and reversible actions, error surfaces and tool result design.
3. 03Context engineering and agent memory9% of the exam

   • Design context and memory strategies that keep agents accurate and efficient over long-running tasks.
   • Scope: Context window budgeting, compaction and summarisation, short- and long-term memory stores, retrieval-augmented generation, hybrid search, reranking and prompt caching.
4. 04Multi-agent orchestration and delegation9% of the exam

   • Evaluate when multi-agent designs outperform single agents and architect orchestration with clear responsibilities.
   • Scope: Orchestrator-worker and hierarchical patterns, sub-agent delegation, agent-to-agent (A2A) protocols, handoffs, shared state and failure isolation.
5. 05Agent evaluation and benchmarking10% of the exam

   • Design evaluation suites that measure agent capability and detect regressions before release.
   • Scope: Task-success and trajectory evaluation, LLM-as-judge calibration, golden datasets, SWE-bench and tau-bench style harnesses, regression suites and statistical significance.
6. 06Agent security and prompt-injection defence10% of the exam

   • Evaluate agent attack surfaces and design layered defences against injection, exfiltration and excessive agency.
   • Scope: OWASP Top 10 for LLM and agentic applications, direct and indirect prompt injection, data exfiltration paths, sandboxing, least-privilege credentials and human-in-the-loop approvals.
7. 07Guardrails, alignment and human oversight8% of the exam

   • Design guardrails and oversight mechanisms that keep agent behaviour within defined policy and risk tolerance.
   • Scope: Policy enforcement layers, input and output classifiers, constitutional and rule-based constraints, escalation and approval gates, autonomy levels and reversibility.
8. 08Observability, tracing and debugging of agents8% of the exam

   • Diagnose agent failures from traces and telemetry and drive targeted fixes to prompts, tools or control flow.
   • Scope: OpenTelemetry GenAI semantic conventions, trace and span analysis, LangSmith and Langfuse, failure taxonomy, replay and root-cause analysis of agent trajectories.
9. 09Model selection, cost and latency engineering8% of the exam

   • Evaluate model and runtime choices to meet quality, latency and cost targets for production agents.
   • Scope: Model routing and cascades, reasoning effort trade-offs, token economics, caching, batching, streaming, structured outputs and rate-limit management.
10. 10Production deployment and agent operations13% of the exam

    • Architect production agent platforms that are durable, versioned and operable at scale.
    • Scope: Durable execution (Temporal), checkpointing and resumability, async task queues, versioning of prompts and tools, canary releases and incident handling for agents.
11. 11AI governance, compliance and risk for agentic systems5% of the exam

    • Lead governance and compliance practices that make agentic systems accountable, auditable and lawful.
    • Scope: EU AI Act obligations, ISO/IEC 42001 AI management systems, NIST AI RMF, data protection, audit trails, accountability and documentation of autonomous actions.

Question types

Single choice 100%

Delivery

Online (remote proctored), Test centre, On-site (corporate)

Retakes

Up to three attempts within any 12-month period. A 14-day waiting period applies after a first unsuccessful attempt and 30 days after a second. Each retake requires a new booking and fee. Retakes draw a different randomised form from the question bank.

Provisional cut score — psychometric validation is required before live delivery. Blueprint v1.0.

1. 1

   Register

   Create a free account and start the programme.

2. 2

   Practice

   Online practice test (coming soon).

3. 3

   Training

   Optional

   Online course or classroom — or skip it.

4. 4

   Certification

   Independent exam and certification decision.

Training is optional and does not by itself confer certification. Find a training partner