Home › Study guides › CCAR-P › Domain 3

CCAR-P · Domain 3 of 7 · 8 lessons · about 173 min

Domain 3: Integration

Connecting Claude to enterprise systems: capability bloat, auth, accuracy vs latency, observability, RAG, retrieval and protocols. 19% of CCAR-P.

This domain is 19% of CCAR-P, about 12 of the 63 questions, the largest on the exam. It is about the joins: how a Claude solution reaches data and systems, with whose permissions, how fast, and how you see what it is doing once thousands of people use it.

The vocabulary: capability bloat is an agent holding more tools or permissions than its role needs. Retrieval-augmented generation (RAG) finds relevant passages in your documents and gives them to the model with the question; chunking is how documents are cut into those passages and indexing is how they are made searchable. The Model Context Protocol (MCP) is an open standard for exposing tools and data to AI applications. Progressive discovery loads a short description of what is available first and the details only when needed, instead of everything up front.

The recurring test: remove the cause, do not watch it. The official sample question has the pattern: an agent that holds refund and delete tools its users never need should lose them, not gain logging or a confirmation step. The same logic runs through the domain: permissions enforced in the system rather than in the prompt, retrieval matched to how the data is shaped, a protocol chosen for who owns each side of the connection.

What the exam guide tests

The official CCAR-P guide lists 8 objectives for this domain. Exam questions are written against them, and so are the lessons: each row says where it is covered.

Lessons

1. 3.1 Capability bloat: auditing an agent's tools and permissionsWhy an agent holding tools its role never needs picks wrong, costs more and widens the attack surface, and how to audit, remove and scope each role's tools.22 min
2. 3.2 Authentication and authorisation: whose identity does each tool call use?Service accounts versus delegated access, the confused deputy, why authorisation lives in the tool and not the prompt, and how to trace every call for gaps.20 min
3. 3.3 Accuracy versus latency: measure both, then justify the configurationHow to measure time to first token, p95 latency and accuracy on one eval set, which levers trade one for the other, and how to justify what you ship.22 min
4. 3.4 Observability at scale: tracing, sampling and scoring LLM trafficWhy an LLM system fails with HTTP 200, what to trace in every request, and how to sample, redact and score millions of conversations at a sane cost.23 min
5. 3.5 Designing a RAG pipeline: chunking, metadata and indexingHow to design a RAG pipeline: chunk by document structure, label every chunk, index for meaning and exact terms, keep the index current, cite sources.22 min
6. 3.6 Retrieval strategies matched to the data and the questionDense, keyword and hybrid search, filters, SQL tools and agentic search: which fits which data and question, why vectors cannot count, when to skip it.22 min
7. 3.7 MCP, API, CLI or agent-to-agent: choosing the integration mechanismHow to choose MCP, a direct API tool, a CLI or agent-to-agent (A2A) for each connection, by ownership, reuse, auth, state, latency and audit.21 min
8. 3.8 Progressive discovery or monolithic context: what an agent loads up frontWhen to load every tool, runbook and document up front, when to let the agent discover detail on demand, and how to design the hybrid in between.21 min

Practice question

From the CCAR-P bank, tagged to this domain. Every answer option is explained. Nothing is stored, nothing to sign up for.