Other certification · CCDV-F
Claude Certified Developer — Foundations (CCDV-F)
Overview guide for the Claude Certified Developer — Foundations exam — the hands-on implementation credential in the Claude certification family. Facts below are taken from the official Exam Guide v1.0 (effective July 2026).
Looking for the Architect exam? The full deep-dive study guide in this repo covers CCAR-F. This page is an overview of a sibling certification — with heavy content overlap, mapped below.
What It Validates
That you can build, integrate, and ship production-grade applications, agents, and workflows on the Claude platform at a foundational level: building agents with the Claude Agent SDK and custom agent loops, integrating Claude through the API and client SDKs (streaming, error handling, multi-format input), operating Claude Code (Skills, CLAUDE.md, settings.json, plugins), prompt and context engineering, designing and running evals, model/cost/latency optimization, secure-by-design practices and hook-based guardrails, and building custom tools and MCP servers.
Intended for: AI/ML engineers, technical leads, and senior software engineers. Recommended (not required): 1–5 years of software engineering, 6+ months hands-on with Claude or comparable LLM systems, Python and/or TypeScript, REST APIs and CLI fluency.
Not intended for: non-technical users, or roles limited to prompt writing without broader application-development responsibility.
Exam at a Glance
| Exam code | CCDV-F |
| Items | 53 (multiple-choice and multiple-response; each item states how many responses to select) |
| Time limit | 120 minutes |
| Delivery | Proctored — online proctored and/or Pearson VUE test center |
| Passing score | 720 scaled (scale 100–1,000) |
| Exam fee | $125 USD |
| Validity | 12 months from award |
| Result reporting | Pass/fail with scaled score, plus percent-correct by domain |
| Prerequisites | None mandatory — recommended experience only |
Domain Blueprint
Unlike the Architect Foundations exam (5 evenly weighted domains), CCDV-F has 8 domains with very uneven weights — a third of the exam is Applications and Integration:
| # | Domain | Weight |
|---|---|---|
| 1 | Agents and Workflows | 14.7% |
| 2 | Applications and Integration | 33.1% |
| 3 | Claude Code | 3.1% |
| 4 | Eval, Testing, and Debugging | 2.6% |
| 5 | Model Selection and Optimization | 16.8% |
| 6 | Prompt and Context Engineering | 11.0% |
| 7 | Security and Safety | 8.1% |
| 8 | Tools and MCPs | 10.6% |
What each domain covers
- Agents and Workflows (14.7%) — workflow-vs-agent decision criteria, manager/supervisor hierarchies, subagents; building with the Claude Agent SDK, custom loops and harnesses, hosted vs self-hosted deployment, hooks for deterministic actions; agent design patterns (tool-use loops, sub-agents, memory, context-window management) and frameworks.
- Applications and Integration (33.1%) — the big one. Requirements and systems-lifecycle basics; Claude API mechanics (messages, tools, streaming, vision, thinking, caching, third-party vendors, batch vs realtime); software-engineering foundations (REST, JSON, async, version control, refactoring); application design (how Claude interprets instructions across Claude Code / Desktop / claude.ai / API / SDKs, content boundaries, schema design, session hygiene); configuration management (CLAUDE.md, settings.json, model pinning, prompt versioning).
- Claude Code (3.1%) — core components (Rules, Skills, Commands, Agents, Agent Memory), session management, slash commands, headless mode, streaming mode, auto-mode, the CLAUDE.md hierarchy, settings.json.
- Eval, Testing, and Debugging (2.6%) — error-type identification, recovery strategies, trace analysis, isolating integration-layer vs model-output problems.
- Model Selection and Optimization (16.8%) — LLM fundamentals (tokens, context windows, sampling, non-determinism), model options (fast mode, extended/adaptive thinking, effort levels), model-tier tradeoffs and breaking changes across releases, token budgeting, prompt caching for cost.
- Prompt and Context Engineering (11.0%) — context-drift/bloat prevention (tool-output pruning, compaction), context isolation via subagents, prompt engineering principles, structured-output patterns, defensive parsing, skepticism toward confident output.
- Security and Safety (8.1%) — prompt-injection awareness and mitigation, jailbreak defense, untrusted input, PII, guardrail layering, hooks as safety controls, secrets and key management.
- Tools and MCPs (10.6%) — tool use and function calling, tool-description writing, client-side vs server-side tools, approval patterns, MCP server authoring and deployment, stdio vs socket transports.
Overlap with This Repo's CCAR-F Material
A large share of CCDV-F content is already covered in the Architect guide, at architect altitude:
| CCDV-F domain | Covered in |
|---|---|
| Agents and Workflows | d1 — Agentic Architecture |
| Tools and MCPs | d2 — Tool Design & MCP |
| Claude Code | d3 — Claude Code Configuration |
| Prompt and Context Engineering | d4 — Prompt Engineering + d5 — Context Management |
| Applications and Integration / Model Selection | Partially in d1/d4; the API-mechanics and cost-optimization depth is CCDV-F-specific |
The developer exam goes deeper on API mechanics, SDK usage, evals, security, and cost optimization than the Architect Foundations blueprint does.
How to Prepare (per the official guide)
- Study the blueprint and self-assess against each skill
- Build and ship at least one real Claude application end-to-end: API integration, an agent loop or Agent SDK build, custom tools or an MCP server, and an eval
- Review official documentation for the Claude API, Agent SDK, Claude Code, and MCP
- Practice cost levers hands-on: model-tier selection, prompt caching, batch processing
Registration and Policies
- Register via the exam's page on the Anthropic Partner Academy (partner-tier discounts apply at checkout), then schedule through Pearson VUE — online proctoring or a test center.
- Cancel or reschedule free of charge up to 48 hours before the appointment (a cancellation in that window is fully refunded — email Anthropic certification support, since cancelling in Pearson VUE alone doesn't trigger the refund); changes within 48 hours forfeit the fee. The v1.0 Exam Guide still says 24 hours; the live Partner Academy policies page is newer.
- Retakes: up to 4 attempts per rolling 12 months, with waiting periods after each failed attempt (14 / 30 / 90 days), per Pearson VUE program policy.
- Renewal: the credential is valid 12 months; renewing on time is free — review what changed since you certified and pass a non-proctored, open-book assessment (retakes allowed; the new 12 months run from your current expiry). If it lapses, the full exam fee applies again. Registration requires a partner-domain email address. Delivery moved to Pearson VUE and digital badging to Credly on 2026-06-30; candidates must be at least 18. Partner-tier discounts apply at checkout — 50% for Select, Preferred, and Global Premier partners, 100% for Global Premier partners through 2026-12-31. The FAQ lists the exam and its prep content in English and Japanese (one registration; you pick the language when scheduling), and a Japanese translation of the v1.0 exam guide was published on 2026-10-08.