CCAR-P Advanced Agentic Architecture Practice Question
A developer is concerned about the high token cost and latency of an agent that has access to 50 different tools. What is the most effective architectural change to optimize this system?
⚠ Common exam trap
Candidates often suggest fine-tuning the model to handle more tools, overlooking that dynamic injection is the standard architectural approach to reduce context bloat and improve performance.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Dynamically inject tools based on intent classification
Model performance and cost are directly impacted by the size of the system prompt and tool definitions. By implementing a dynamic tool selection mechanism, the architect can ensure that only the tools relevant to the user's current intent are loaded into the context, significantly reducing the prompt overhead for every turn.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Dynamically inject tools based on intent classification
Why this is correct
Using a lightweight classifier to identify the user's intent allows the system to provide only a small subset of the 50 tools. This reduces the input token count, lowers latency, and improves the model's accuracy by removing irrelevant tool schemas that could cause confusion or lead to hallucinations.
- ✗
Switch from JSON to XML tool definitions
Why it's wrong here
While Anthropic models are proficient at parsing both XML and JSON, switching formats does not address the underlying issue of having too many tools in the context window. The token savings from a different format would be negligible compared to the impact of reducing the total number of tools provided.
- ✗
Compress the tool descriptions into single words
Why it's wrong here
Vague or overly compressed tool descriptions degrade the model's ability to understand when and how to use a tool. This leads to increased errors and failed calls, which actually increases costs in the long run as the agent requires more turns to correct its mistakes and successfully complete the task.
- ✗
Force the model to use only one tool per turn
Why it's wrong here
Limiting the number of tools used per turn does not reduce the size of the prompt if all 50 tool definitions are still included in the input. The model still has to process the full set of schemas to make its single selection, so the latency and token costs remain high.
About these practice questions
One of 262 original CCAR-P practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Anthropic exam blueprint
This CCAR-P practice question is part of Courseiva's free Anthropic certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the CCAR-P exam.