CCDV-F vs CCA-F: Which Claude Certification Should You Take? [2026]

Anthropic's certification program grew into four role-based credentials on July 23, 2026, and the expansion created a genuinely new decision. Before that date, a practitioner picking a Claude certification picked a tier. Now two Foundations-level exams sit at the same $125 price and the same 120-minute proctored format, and they point at different jobs. The Claude Certified Developer - Foundations (CCDV-F) certifies engineers who build and ship applications on the Claude API, the developer platform for calling Claude models programmatically. The Claude Certified Architect - Foundations (CCA-F), the original exam in the program from March 2026, certifies practitioners who design agent systems: choosing orchestration patterns, defining tool interfaces, and keeping the result reliable in production.

Where our CCA-F vs CCAR-P comparison contrasts two altitudes of the same architect track, this comparison contrasts two roles at the same altitude. That changes the shape of the choice: instead of asking how senior you are, you ask what kind of work fills your week. This guide puts the two blueprints side by side, maps where they overlap, and closes with a decision framework by role and career goal. For the individual deep dives, see the complete CCDV-F guide and the complete CCA-F guide.

At a Glance

The format family is identical by design. Both exams register through the Anthropic Partner Academy, Anthropic's training portal on Skilljar, which requires joining the Claude Partner Network (free at claude.com/partners). Both deliver through Pearson VUE with identity verification, either online proctored or at a physical test center. Both score on a scaled 100-1000 range where 720 passes: scaled scoring converts raw correctness across exam forms of varying difficulty onto one comparable scale, and your score report shows percent-correct per domain. Both issue a digital badge through Credly by Pearson, stay valid for one year, and renew through a free non-proctored assessment on the Partner Academy, provided you complete it before the 12-month expiration.

The structural differences are the question count (53 against 60, which gives CCDV-F slightly more time per item) and the blueprint shape. CCDV-F spreads its weight across eight domains with one dominant third, while CCA-F concentrates everything into five domains at the design layer. That shape difference is the whole story, so the next section walks through it.

What Each Exam Actually Tests

CCDV-F: Application Mechanics From Request to Production

CCDV-F is an application engineering exam, and one number defines its character: the Applications and Integration domain carries 33% of the blueprint, roughly double the heaviest domain on most certifications. It covers the Messages API end to end (requests, tool calls, streaming, vision inputs, extended thinking), prompt caching (reusing a stored prompt prefix so repeated calls cost less and return faster), the choice between batch and realtime processing, and configuration management across CLAUDE.md files, settings.json, model pinning, and prompt versioning. If you have shipped a Claude-backed feature, most of that list reads like your recent commit history.

The next two domains extend the same theme. Model Selection and Optimization (17%) tests LLM fundamentals (tokens, context windows, sampling, non-determinism) plus the practical trade-offs among Opus, Sonnet, and Haiku and the cost levers a production budget depends on. Agents and Workflows (15%) covers when a deterministic workflow beats an autonomous agent, manager and subagent hierarchies, and building with the Claude Agent SDK, Anthropic's toolkit for constructing agents with custom loops and hooks, alongside frameworks like Strands, LangGraph, and PydanticAI. The remaining five domains form a long tail: Prompt and Context Engineering (11%), Tools and MCPs (10%), Security and Safety (8%), and slim 3% slices for Claude Code and for evaluation, testing, and debugging. The full CCDV-F domain breakdown walks every topic with study guidance.

CCA-F: Design Judgment Over the Same Platform

CCA-F concentrates its five domains on the design layer. Agentic Architecture and Orchestration (27%) anchors the exam: designing agentic loops (the cycle where Claude plans an action, calls a tool, observes the result, and decides whether to continue or stop) with correct termination conditions, hub-and-spoke orchestration, and subagent isolation. Claude Code Configuration and Workflows (20%) goes deep on Anthropic's agentic coding tool, reaching into CLAUDE.md hierarchies, custom skills, permission hooks, and CI/CD integration. Prompt Engineering and Structured Output (20%) covers few-shot design and constraining responses to a machine-parseable format like JSON that downstream code can validate. Tool Design and MCP Integration (18%) tests how you specify tool interfaces and wire the Model Context Protocol (MCP), the open standard for connecting Claude to external tools and data sources, without leaking secrets. Context Management and Reliability (15%) closes with token budgets, degradation effects like lost-in-the-middle, and recovery from failures mid-session.

Notice what that list assumes. API mechanics, model economics, and SDK construction barely appear as named topics because CCA-F treats them as background you already carry. Its questions are scenario-based and production-flavored: a loop without an exit condition to spot, a tool description that misleads the model, a context strategy that degrades as a session grows.

One Protocol, Two Question Styles

MCP illustrates the role split cleanly. CCDV-F asks builder questions about it: implement a server exposing resources, tools, and prompts over a stdio transport, and decide when a built-in tool, a custom tool, a Skill, or a full MCP server is the right packaging for a capability. CCA-F asks designer questions about the same protocol: write the tool description an orchestrated agent can actually use, pick the right configuration scope for a team, and handle the structured error a server returns mid-loop. Preparing for one teaches you how the protocol works, and preparing for the other teaches you what to do with it.

Where They Overlap (and Where They Diverge)

Four topics transfer almost directly between the blueprints. Prompting technique, context engineering (the discipline of deciding what enters the model's context window and in what order), agent fundamentals, and MCP concepts appear on both sides, and together they touch close to half of each exam. A candidate who studies either blueprint honestly arrives at the other one with a running start.

The divergence is just as concrete. CCDV-F places 50% of its weight (Applications and Integration plus Model Selection) on material CCA-F never names as a domain, and it holds the only dedicated Security and Safety domain of the pair, covering prompt-injection defense, data leakage, and guardrail layering. CCA-F answers with a 20% Claude Code domain against the 3% CCDV-F slice, and it takes the shared topics further up the stack: CCDV-F asks you to prevent context bloat in your own application, while CCA-F asks you to design the context strategy an entire agent system depends on. The altitude within each shared topic differs even when the topic name matches.

Which Should You Take? A Decision Framework

You Are an Application Engineer Shipping Claude Features

Take CCDV-F. The blueprint mirrors the job: half the exam covers the API calls, model choices, and cost controls you handle in every sprint, and the Security and Safety domain matches the review checklist your features already pass through. The recommended profile (1-5 years of software engineering, working Python or TypeScript, REST and CLI fluency, six months with Claude or a comparable LLM stack) describes a working product engineer rather than a specialist. Start with the how to pass CCDV-F guide to see what first-attempt preparation looks like.

You Are a Solution Architect or Platform Engineer

Take CCA-F. Your deliverable is the design: which orchestration pattern fits the problem, where the tool boundaries sit, how context is budgeted, and how Claude Code is configured across a team. CCA-F examines exactly that judgment, and its 47% combined weight on orchestration and Claude Code workflows rewards the system-level fluency your role builds daily. It also positions you for the Professional tier, since the architect track continues upward through CCAR-P.

You Are Moving From General Software Engineering Into AI

Start with CCDV-F. Its entry assumptions match the background you already have, and its bottom-up sequence (API mechanics first, then models, then agents) teaches the platform in the order you will actually use it. The 4-week CCDV-F study plan is paced for exactly this profile. Add CCA-F later, once design decisions start landing on your desk, and the overlap will make that second preparation noticeably shorter.

Your Goal Is the Architect Track and Eventually CCAR-P

CCA-F is the straighter line. Its domains feed directly into the CCAR-P blueprint, and holding it signals commitment to the architecture path. CCDV-F still helps (the ladder section below shows how), so treat it as a strengthening move rather than a required rung.

Taking Both: The Order That Makes Sense

For most engineers the sequence is CCDV-F first, then CCA-F, because mechanics precede design. Understanding how streaming, caching, tool calls, and model selection behave in real applications gives the CCA-F design scenarios something to stand on, and the shared ground in prompting, context, agents, and MCP means the second preparation concentrates on genuinely new territory: orchestration depth and the 20% Claude Code domain. The reverse order suits practicing architects who need the design credential now and want to backfill documented implementation depth later.

The pair costs $250 in first-attempt fees, and the ongoing cost stays low because both renew through the free annual assessment on the Anthropic Partner Academy. Holding both tells a hiring manager you can build the application and design the system it lives in, which is a rare combination on paper even when it is common in practice.

How Both Ladder Toward CCAR-P

The Claude Certified Architect - Professional (CCAR-P, $175, 63 questions, seven domains) sits above both Foundations exams and recommends three or more years of systems architecture experience. Each Foundations credential feeds it differently. CCA-F maps onto the CCAR-P Solution Design and Architecture domain (17%) and its Developer Productivity and Operational Enablement domain (7%), and its MCP depth supports the heavyweight Integration domain (19%). CCDV-F reinforces that same Integration domain from the implementation side and covers the ground under Claude Models, Prompting and Context Engineering (13%), while its Security and Safety domain previews part of Governance, Safety and Risk Management (14%).

Neither Foundations exam touches the CCAR-P territory that most candidates find hardest: Evaluation, Testing and Optimization (16%) and Stakeholder Communication and Lifecycle Management (14%) reward lived architecture experience that no Foundations blueprint can substitute for. Plan for those domains separately whenever you make the jump; the CCA-F vs CCAR-P comparison covers that transition in detail.

Frequently Asked Questions

Choose, Then Calibrate With a Practice Test

The comparison gives you the decision, and a timed practice exam turns it into a plan. A baseline score tells you whether your chosen exam is three weeks away or eight, and the per-domain results show exactly where the work is.

Preporato covers both tracks with domain-weighted practice exams. The CCDV-F practice tests include six full-length exams of 53 questions each, mirroring the eight-domain blueprint with the same multiple-response mix as the real exam, plus a 500-card flashcard deck for the definitional layer. The CCA-F practice tests provide the same treatment for the five architect domains. Both sets come with explanations for every answer and are available together through Preporato Pro, so taking both certifications never requires a second purchase.

Sources:

• Anthropic four-certification announcement
• Anthropic Partner Academy
• Claude Developer Documentation

Last updated: August 8, 2026