Internal Claude projects that build real experience without touching customer data
This page on internal Claude projects covers hands-on practice for certified teams on a firm’s own non-sensitive material, and Cred Farmer’s six, each mapped to a named v1.0 guide domain.
Short answer
Internal Claude projects can stay clear of customer data by using only a firm’s own non-sensitive material: its handbook, public site and own code. Cred Farmer’s six each map to a named domain of a v1.0 exam guide, because every guide lists building something under how to prepare. On Cred Farmer’s reading of Anthropic’s announcement, a firm’s count asks that its people have “used Claude in the past 90 days”, so use is a standing question. No official page Cred Farmer could read says a project satisfies that test.
Cred Farmer is independent and is not part of the Claude Partner Network or an Anthropic source. The facts come from the official pages named below, read on 6 November 2026, and Anthropic’s partner team, not this page, decides a firm’s standing. Cred Farmer also publishes practice papers, so it has a stake here.
What the exam codes mean
- CCAO-F
- Claude Certified Associate - Foundations
- CCDV-F
- Claude Certified Developer - Foundations
- CCAR-F
- Claude Certified Architect - Foundations
- CCAR-P
- Claude Certified Architect - Professional
Which internal projects can help our certified team gain practical experience without exposing sensitive customer data?
Six internal projects can give a certified Claude team practical experience without customer data, on Cred Farmer’s list. They are a handbook Project, a public-page checker, a site answer app, a repository set-up, a site-review agent and a retrieval service. No official page Cred Farmer could read sets such a list, so its limits are Cred Farmer’s.
A project here means internal work. Capital-P Projects is a Claude feature, which only the first needs. The four guides are for Claude Certified Associate - Foundations (CCAO-F), Claude Certified Developer - Foundations (CCDV-F), Claude Certified Architect - Foundations (CCAR-F) and Claude Certified Architect - Professional (CCAR-P).
| Project | Firm’s own material | Guide v1.0 and domain | Weight |
|---|---|---|---|
| Handbook Project | Handbook, personal details removed | CCAO-F, Domain 5: Configuration and Knowledge Management | 12% |
| Public-page checker | Public site | CCAO-F, Domain 2: Output Evaluation and Validation | 21%, heaviest |
| Site answer app | Public site | CCDV-F, Domain 2: Applications and Integration | 33.1%, heaviest |
| Repository set-up | Own code, no client code or credentials | CCAR-F, Domain 3: Claude Code Configuration & Workflows | 20% |
| Site-review agent | Public site | CCAR-F, Domain 1: Agentic Architecture & Orchestration | 27%, heaviest |
| Retrieval service | Handbook and public site | CCAR-P, Domain 3: Integration | 19%, heaviest |
Heaviest is Cred Farmer’s reading, and the guides call weights approximate. Two projects map to CCAO-F, which Anthropic’s FAQ lists among exams that do not count towards Claude Partner Network eligibility. Cred Farmer keeps them because that guide’s audience includes “internal staff who maintain and optimize ongoing AI-enabled workflows”.
Claude practice projects for teams start from the guides’ build lines
Claude practice projects for teams have a published starting point, because each v1.0 exam guide lists something to build under how to prepare. The guides write those lines for candidates. Carrying them into a certified team’s year is Cred Farmer’s step.
- CCAO-F: “Build real workflows: configure a Project with instructions and knowledge sources”
- CCDV-F: “Build and operate at least one Claude application that exercises the API, integrates one or more tools, applies basic prompt and context engineering, and includes simple security and evaluation practices”
- CCAR-F: “Build an agent with the Claude Agent SDK” and “Configure Claude Code for a real project”, plus four numbered exercises
- CCAR-P: “Build and operate at least one end-to-end Claude solution, including RAG, evaluation, and observability”
Anthropic’s FAQ says “Start with the exam guide for your certification”, and Partner Academy’s Associate path ends on “Take one workflow from your own job”.
Anthropic’s certifications page says CCAO-F, CCDV-F and CCAR-P will include updated content from 7 December 2026, with updated guides due on 17 November. That notice omits CCAR-F. This map is v1.0, so re-read it for a test on or after 7 December: see Claude certification changes.
Six internal Claude projects, one domain each
Each of Cred Farmer’s six practice projects for a certified Claude team ends in something a colleague can review: a log, a test set, a set-up or proposed edits. All six are Cred Farmer’s advice, not an official list.
Handbook Project
Load the handbook, personal details removed, into a Claude Project with written instructions. Log the staff questions it gets wrong. When a page changes, update it and ask again.
Public-page checker
Have Claude rewrite a service page for a new audience, then trace each claim back to the page. Log every detail the page does not print.
Site answer app
Through the API, answer questions about the firm’s services from its public pages, with one page-lookup tool, a system prompt and a short test set. Treat fetched text as untrusted input.
Repository set-up
On a repository the firm owns, set up what the CCAR-F guide lists: CLAUDE.md, path-specific rules, a custom skill and one MCP server. Have a colleague test it from a fresh session.
Site-review agent
On the Claude Agent SDK, build an agent that reads the public site through tools and proposes edits where it disagrees with the service list. A hook blocks publishing.
Retrieval service
Answer staff questions from the handbook and public site through retrieval. Compare chunking choices on a test set, and log each answer. CCAR-P v1.0’s Domain 4, at 16%, covers that testing and logging.
What stays out: Cred Farmer’s data limit, not an Anthropic rule
Cred Farmer’s data limit for its six practice projects is not an Anthropic rule. The list uses the firm’s handbook, public site and own code, leaving out customer records, client files and credentials. Exam content has Anthropic’s own rule. CCAO-F’s Domain 6, Governance, Risk, and Responsible Use, at 15% in Exam Guide v1.0, lists “data sensitivity, regulatory, and privacy considerations”.
Before a file goes in
That domain also lists “organizational AI policies”, the third check here, and governance and responsible use on CCAO-F covers how it is examined. Partner Academy’s course of that name, 55 minutes, is listed as “Which data is safe to share where, and the judgment calls of everyday use”.
Anthropic’s pricing page, read on 6 November 2026, lists model training as Opt-out on Free and Pro, and None by default on Team and Enterprise. Which plan a firm uses is its own choice, and a Claude practice environment has the prices.
Anthropic’s Commercial Terms and privacy documents are for a lawyer, and this page states no rule from them. Your firm’s confidentiality duties and data law go to a lawyer in your country.
Exam Policy section 1 says “Exam Content is considered Anthropic’s Confidential Information”. Whether pasting a course into a Project falls under Certification Terms section 7(b), which limits how “Program content” is copied or stored, is a lawyer’s question. So Cred Farmer’s six leave courses out: see internal bootcamps.
The 90-day line makes use a standing question, on Cred Farmer’s reading
Claude use is a standing question on Cred Farmer’s reading of Anthropic’s announcement of 3 June 2026. It describes the certified count as people who “hold a current Anthropic certification and have used Claude in the past 90 days”. No official page Cred Farmer could read says an internal project satisfies that test.
That reading joins two lines printed apart: the tier lines say “active certified individuals”, and a later line gives the two tests. The FAQ adds the exam list and the email domain. No page read defines used Claude or says how it is measured: see what an active certified practitioner is.
So this page does not offer a project as a way to keep anyone in a count. That question goes to your partner contact or partner-support@anthropic.com. The announcement says of “The best partners” that “They use the newest models for their own work before they put it in front of a client”, and does not tie that line to the count.
Cred Farmer’s line: a project is practice, not a count
No page read says a project counts as having used Claude. With no customer in it, it adds nothing to the customer and story counts, on Cred Farmer’s reading.
Customer deployments covers those counts, and ready for client work covers the step from a pass to a client project.
Who takes which project, and how long it runs
Cred Farmer’s advice is to give each certified person the Claude practice project built on their own exam’s guide, so a CCDV-F holder takes the site answer app. No official page it could read requires internal projects, or sets their hours or cadence.
The Claude Code partner badge path, “built for consultants”, estimates “~10-15 hours” for exercises and artifacts, and Partner Basecamp is “Two days of hands-on technical training with Claude for partner teams”.
Cred Farmer’s advice is to re-run each project when Anthropic ships a new model, not on a 90-day clock. The CCDV-F v1.0 guide lists “breaking behavior changes across model releases”.
A project buys no right to sit an exam: see the Claude certification eligibility guide. Fees are in the cost guide, and the hours belong in the team’s business case.
Key takeaways
- Cred Farmer’s six practice projects use only a firm’s own non-sensitive material: its handbook, public site and own code.
- By Cred Farmer’s reading, four of its six projects sit on a v1.0 guide’s heaviest domain. They run from 19% on CCAR-P to 33.1% on CCDV-F.
- No official page Cred Farmer could read on 6 November 2026 says an internal project counts as having used Claude. The Partner Hub could not be read, so ask your partner contact.
- Cred Farmer’s data limit is not an Anthropic rule. A firm’s confidentiality duties and data law go to a lawyer in its own country.
A project uses Claude. A mock rehearses the clock.
Cred Farmer’s free CCDV-F mock is a question paper with no Claude in it. It follows Exam Guide v1.0, for a sitting up to 6 December 2026. A free sitting saves no result and forecasts nothing about the real exam.
Sit the free CCDV-F mock: timed, no account neededFrequently asked
Does Anthropic offer Claude sandbox projects for partner firms?
No official page Cred Farmer could read on 6 November 2026 offers partner firms a sandbox. The Partner Hub, Program Guide and Partner Agreement could not be read, so ask your Anthropic contact. The nearest published material Cred Farmer found is the exercises in the CCAR-F guide and the Claude Code badge path.
Does an internal project count as having used Claude in the past 90 days?
No official page Cred Farmer could read on 6 November 2026 says. Anthropic’s announcement of 3 June 2026 gives those words and no test of use. The Partner Hub could not be read, so ask your partner contact or partner-support@anthropic.com.
Can hands on Claude experience for consultants come from internal work?
No v1.0 exam guide says that experience must come from client work. The CCAO-F guide recommends “Regular, hands-on experience using Claude in a professional setting”, and the CCAR-P line adds “in production”, a term it does not define. Both call those lines recommended, not required.
Will these projects help someone pass a Claude exam?
Cred Farmer promises nothing of the kind. Three v1.0 exam guides say “Anthropic does not guarantee that any particular resource ensures a passing result”. No official page read, nor any Cred Farmer figure, measures what a project does for a score.
Related posts
Not affiliated with, or endorsed by, Anthropic or Pearson VUE. Details are summarised from publicly published program information and can change. Always confirm against the official exam guide before booking.