By Amotion AITimoLabs › Claude Certification › Developer Foundations
CCDV-FClaude Certified Developer Foundations (CCDV-F)
CCDV-F is the Claude Certified Developer – Foundations certification for people who build Claude applications, tools and agents.
CCDV-F quick facts
| Questions | 53, multiple choice and multiple response |
|---|---|
| Time | 120 minutes |
| Exam fee | US$125 |
| Pass mark | Scaled score of 720 on a 100 to 1,000 scale (not 72%) |
| Valid for | 12 months from the award date |
| Delivery | Proctored, online or at a test centre |
| Prerequisites | None. Recommended: 1 to 5 years of software engineering and 6+ months with Claude or comparable LLM systems |
| Your result | Pass or fail, a scaled score, and percent correct by domain |
What the exam covers
| 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% |
Applications and Integration is a third of the exam. Claude Code and evaluation together are under 6%.
Try the free practice exam. Timo membership includes five full CCDV-F mock exams. To register, you need a Claude Partner Network affiliation; see how to register without a partner employer.
Source: Anthropic CCDV-F exam guide, version 1.0, effective July 2026. official Developer Foundations certification route.
Anthropic’s listed exam fee is before any partner-tier discount. See the Claude certification cost guide for official fees and payment details.
Prepare for CCDV-F with Timo
Timo's Developer Foundations guide helps you connect exam topics to the work of building a Claude application. You can use the public study domains, implementation exercise and free practice exam to find gaps in your preparation.
Amotion AI, which runs Timo, is Certified in the Claude Partner Network. If your employer is not a Claude partner, you can enroll as a consultant with Timo by Amotion AI, get certified as a Claude Certified Architect or Developer, and work on enterprise AI projects. Enrolled consultants sign a consulting agreement with Amotion AI, which is Certified in the Claude Partner Network, register for their Claude exam with their Amotion AI consultant email, and use their expert profile and Timo Scout AI to find enterprise AI projects. Anthropic's official exam registration and fee are separate. Read the application steps.
Apply for partner learning accessWho should use this guide
Use this route if you implement Claude inside a product or internal system. The role requires more than prompt writing. You turn a requirement into an application that calls Claude, checks the data it receives, limits what tools may do and handles failures after release.
A practical starting point is a small service written in Python or TypeScript. It should accept a request, call Claude, validate structured output and save an evaluation record. This makes every study topic concrete. Candidates comparing roles can start at Claude certification courses.
Developer study domains
Agents and workflows
Choose a fixed workflow when steps are predictable. Use an agent loop when the task needs iterative decisions. Define a stopping condition, approval boundary and state passed between steps. A loop without limits can repeat work and consume tokens.
Applications and integration
Connect the API to an actual application interface. Handle request validation, streaming where appropriate, asynchronous work, configuration, version control and deployment. Keep the model boundary clear inside the wider service.
Claude Code
Use project instructions, permissions and review habits that make coding sessions repeatable. Treat generated changes as code requiring tests and human review. Repository guidance should define relevant conventions and safe limits.
Evaluation, testing and debugging
Build a representative test set before release. Compare outputs against a stated rubric, inspect failures and record the input, model setting and result. A happy-path demonstration gives little evidence about difficult cases.
Model selection and optimisation
Choose a model and request pattern for the task. Measure quality, latency and cost together. Manage tokens and context deliberately, then use caching where it supports the intended experience.
Prompt and context engineering
Write clear instructions, provide relevant context and specify the output shape. Separate trusted instructions from untrusted content. Version prompts so an evaluation result links to a concrete implementation change.
Security and safety
Keep secrets outside prompts and repositories. Apply least privilege to data and tools. Treat prompt injection, malformed requests and unsafe actions as system-design concerns with technical controls and human escalation.
Tools and Model Context Protocol
Give each tool a narrow purpose, precise schema and useful error response. MCP can expose reusable tools and resources across clients. Its connections need authentication, permission boundaries and observable failure behaviour.
CCDV-F study guide: all 25 topics
One page per official topic, each with two sample questions and a link to the free practice exam.
Domain 1: Agents and Workflows (14.7%)
Domain 2: Applications and Integration (33.1%)
Domain 3: Claude Code (3.1%)
Domain 4: Eval, Testing, and Debugging (2.6%)
Domain 5: Model Selection and Optimization (16.8%)
Domain 6: Prompt and Context Engineering (11.0%)
Domain 7: Security and Safety (8.1%)
Domain 8: Tools and MCPs (10.6%)
Developer Foundations practice questions
This is independent TimoLabs preparation material. It is separate from Anthropic assessment content. Use it to identify the build evidence you still need.
- Explain why a workflow or agent loop fits one user request.
- Show how your API handler rejects invalid input early.
- Write a tool schema with an approval boundary and useful error response.
- Show how tool results return safely to conversation context.
- Configure Claude Code guidance and review a generated change.
- Describe the client, server and trust boundary in an MCP connection.
- Evaluate failures using a labelled test set and stated rubric.
- Trace latency, token use and tool failures during one request.
- Protect secrets and constrain an action with least privilege.
- Choose a retry, fallback or human hand-off for a reliability failure.
Mark every item with evidence, a code link or a short architecture note. Turn weak answers into the next build task. The point is a defensible decision record, not a score.
CCDV-F practice: turn a support request into one validated work item
Developer Foundations preparation should show how a model response becomes a reliable application result. A successful API call is only one part of that process.
This original Timo exercise uses invented data. A support team wants a small application that reads and classifies a submitted request, then drafts a support ticket for a human support agent to review. The application stores these work items in its own database and must not create duplicates when someone submits the same request twice.
Define the application result before calling Claude
Use a request ID supplied by the application, such as REQ-204, and sample text: “My invoice shows two charges for the same order. Please check it.” The expected result contains that request ID, a category such as billing, a summary and a review status. The application owns the request ID; Claude should not invent or replace it.
Ask Claude to classify and summarise the request using the defined output structure. Keep the original text alongside the draft so the support agent can check what the customer actually said. The application should describe the report as a possible duplicate charge, not a confirmed billing error: classification does not establish that the charge was wrong.
Handle failure before creating a work item
Validate the completed model response before creating the work item in the application's database. If the response is incomplete or cannot be accepted, record the failure against REQ-204 and route it for retry or manual review according to the application's policy. Do not create a partly filled work item just because the API returned a response.
For an accepted draft, use REQ-204 as the application's duplicate-prevention key. Create the record with an atomic database insert protected by a unique request-ID constraint, so two simultaneous submissions cannot both create a record. If a record already exists, return it rather than create another. This control belongs in application code and storage; asking Claude not to duplicate work cannot enforce it across requests.
For a tool action that changes a business record, work through the Claude Agent SDK hooks example to see where an application can inspect and block a proposed call.
Test duplicate submissions and interrupted responses
Submit the same request twice and confirm that it resolves to one work item. Submit an incomplete or ambiguous request and confirm that the application preserves the uncertainty for the reviewer. Simulate an interrupted model response and confirm that no incomplete work item is created.
The exercise is complete when you can show the input, validated draft, resulting record and failure log, and explain which component owns each decision. It connects application integration, structured output, error handling, evaluation and operational reliability without requiring an unnecessarily complex agent.
Use the free practice exam to practise choosing an implementation fix from several plausible options. Review the assumption behind an incorrect answer, then test that assumption in your own application.
A study sequence for developers
Map the official domains
Read Anthropic's current guide and create a domain checklist. Link each topic to a file, test or design note from your build.
Build a narrow application
Choose one task. Add a request boundary, Claude call and structured response. Begin without tools so output behaviour stays easy to inspect.
Add tools and an agent loop
Introduce one approved tool, validate arguments and log failures. Add an agent loop when the task needs iterative decisions.
Set an MCP boundary
Document the client, server, allowed resource and credential treatment. Review permissions before testing the connection.
Evaluate and operate
Create normal, ambiguous and unsafe cases. Review quality, latency, cost, observability and recovery. Then use registration when ready.
Preparing in India
Candidates in India follow the same official Developer Foundations certification route. Check Anthropic’s current eligibility, exam format and fees before booking.
Choose online or test-centre arrangements only from the current official registration flow. Compare public learning resources by their scope, update date and treatment of original work. Avoid material that claims access to live exam questions.
Preparing in the United States
Candidates in the United States can start with Anthropic’s official learning route, then use community examples to practise implementation decisions. Use original exercises and explanations when studying; avoid sources that reproduce protected exam content.
For an employer-facing portfolio, show the implementation boundary, tool permissions and evaluation result. A repository with a clear README is easier to review than disconnected prompts.
Developer Foundations FAQ
Who is CCDV-F for?
CCDV-F suits developers who build or integrate Claude applications, agents, tools and workflows. It is a technical route for implementation decisions.
Is coding required for CCDV-F?
The developer track covers APIs, tools, integrations and software delivery. Practical coding work provides the clearest preparation for these decisions.
Which tools should I practise with?
Practise API requests, schema validation, logging, tests and source control. Add a scoped tool when the task needs external data or an action.
Why does MCP matter?
MCP connects AI clients with tools, resources and prompts. Study its interface, authentication and trust boundary as part of a system design.
Are the diagnostic and practice material official?
No. The diagnostic is independent TimoLabs material. It contains no official questions, dumps or reconstructed assessment content.
How can Timo help me prepare for the CCDV-F implementation topics?
Timo provides this Developer Foundations guide and its support-request exercise to help you practise the implementation topics. Then answer the free practice exam and review why each alternative succeeds or fails. Timo enrolls individuals who want to become certified AI consultants and work on enterprise AI projects. Enrolled consultants sign a consulting agreement with Amotion AI, which is Certified in the Claude Partner Network, register for their Claude exam with their Amotion AI consultant email, and use their expert profile and Timo Scout AI to find enterprise AI projects. The official exam is registered and paid for separately through Anthropic's route.
How should I use a CCDV-F mock result to improve my application skills?
Turn each missed concept into a check you can demonstrate. If you chose an unsafe retry, show how your application prevents a duplicate action. If you missed an output-handling question, test an incomplete response and confirm that it cannot reach the database writer. Keep a record of what failed and what changed. A mock result guides revision; it does not predict your official scaled score or guarantee a pass.
Where are official resources?
Use Anthropic's Partner Certifications page and confirm the current guide before registration. Use the Claude certification questions and answers for independent preparation.
What is CCDV-F?
CCDV-F is the Claude Certified Developer – Foundations certification for people who build Claude applications, tools and agents.
Does CCDV-F count toward Claude Partner Network tiers?
Yes. CCDV-F is one of the three certifications that count toward Claude Partner Network eligibility.
Sources and change log
- Anthropic, Partner Certifications, official public listing. Checked 2026-09-27.
- Anthropic CCDV-F official exam guide (PDF), version 1.0, effective July 2026. Exam format, fee, passing score, validity and domain weights checked 27 September 2026.
- TimoLabs market research, India and United States split, dated 2026-08-12.
Reviewer: Poorna Reddy. Last verified: 2026-09-27.
Content updated 20 September 2026: added an original support-request implementation exercise, linked track practice questions and clarified the partner learning application route. Official exam facts retain their source-check dates above.
Ready to apply
Practise the Developer Foundations topics against a small working application. Enroll as a consultant with Timo to prepare for Claude certification and find enterprise AI projects.
Apply for partner learning accessAnthropic controls official exam eligibility, scheduling and credential awards. Review the Timo consultant enrollment process before enrolling. Timo consultants register for Claude certification with their Amotion AI consultant email after completing enrollment and the consulting agreement. Anthropic's exam fee and credential requirements remain separate.
Latest Claude jobs
221 open jobs ask for Claude skills. Updated 2 Nov 2026.
- Research Engineer / Scientist, Pre-trainingAnthropic
LLMs and generative AIPythonPyTorch, TensorFlow and JAXKubernetes and Docker
- AI EngineerThesis
$160k to $200k a year
AI agentsLLMs and generative AIRAG and retrievalMCP
- Director, GTM AI EngineeringSamsara
AI agentsLLMs and generative AIMCPEvaluations
- Engineering Manager, Conversational AIPebl
CA$154k to CA$195k a year
AI agentsLLMs and generative AIEvaluationsPrompt engineering
- AI & Innovation Lead, G&ASanity
$165k to $180k a year
EvaluationsSalesforce and AgentforceClaude and Anthropic