A terminal tool that downloads ExamTopics practice questions for any certification exam.
You type an exam code like SAA-C03. It finds the exam, collects every public discussion
page for it, and writes two plain text files: the links, and the questions themselves with
all answer choices, the suggested answer, and how the community voted.
Covers 2252 exams across 186 providers on ExamTopics — AWS, Microsoft, Cisco, CompTIA,
Google, Fortinet, Salesforce, and the rest. Nothing is hardcoded, so new exams show up on
their own. (ExamTopics lists 188 providers, but elastic and fsmtb have no exams
published on the site at all.)
Plain HTTP only. No browser, no headless runtime, no captcha solver, two dependencies.
A <exam> dumps.txt file full of entries like this:
======================================================================
Topic 1 | Question 3
https://www.examtopics.com/discussions/github/view/321807-exam-github-actions-topic-1-question-3-discussion/
As a developer, which workflow steps should you perform to publish an image
to the GitHub Container Registry? (Choose three.)
A. Use the actions/setup-docker action
B. Authenticate to the GitHub Container Registry.
C. Build the container image.
D. Push the image to the GitHub Container Registry
E. Pull the image from the GitHub Container Registry.
Suggested answer: ABD
Community answer: BCD [BCD 3 (100%)]
Suggested answer is ExamTopics' own answer. Community answer is what commenters voted
for, with the vote split. They disagree often, which is useful to see.
You need Python 3.8 or newer. That is all. (Developed and tested on Python 3.13.)
To check whether you already have it, open a terminal and run:
python --version
If you see something like Python 3.13.1, you are ready — skip to Step 2.
If you see an error, or a version below 3.8, install Python first:
- Go to https://www.python.org/downloads/
- Download the latest version for your operating system.
- On Windows, tick "Add python.exe to PATH" on the first screen of the installer. This is easy to miss and everything else depends on it.
- Finish the installer, then close and reopen your terminal.
Either clone it with git:
git clone https://github.com/swarnava-dutta/Free-Exam-Dumps.git
cd Free-Exam-Dumps
Or, if you do not have git: click the green Code button on the GitHub page, choose Download ZIP, then right-click the downloaded file and Extract All. Open the extracted folder.
Double-click:
install_dependencies.bat
A black window opens and does three things: creates a private Python environment in a
.venv folder, installs requests and tqdm into it, and stops. It leaves the rest of
your computer's Python untouched.
It takes about 25 seconds. When you see this, it worked:
============================================================
Done. Next: double-click run_scraper.bat
============================================================
Press any key to close the window. You never need to run this again unless you move the
folder or delete .venv.
Double-click:
run_scraper.bat
On the very first run you will see a progress bar labelled Loading exam index. The tool
is reading the exam list for all 188 provider pages so it can recognise any code you type.
This takes about 30 seconds, once. The finished list is saved to
.examtopics_index.json (about 160 KB) and reused, so every later run starts instantly.
Exam code or name (example: SAA-C03, AZ-104, SY0-701): saa-c03
Press Enter. The tool confirms what it found before doing any work:
Exam: AWS Certified Solutions Architect - Associate SAA-C03
Provider: amazon
Published: 1019 questions
Scanning: 604 discussion pages
Then it runs two passes, each with a progress bar and a live count:
Scanning pages: 79%|#######9 | 480/604 [00:02<00:00, found=1019]
Wrote aws-certified-solutions-architect-associate-saa-c03 links.txt (1019 links)
Fetching questions: 100%|##########| 1019/1019 [01:34<00:00, 10.77q/s]
Wrote aws-certified-solutions-architect-associate-saa-c03 dumps.txt (1019 questions)
Both files land in the project folder, next to main.py:
| File | Contents |
|---|---|
<exam> links.txt |
every discussion URL, grouped by topic, in question order |
<exam> dumps.txt |
the questions, choices, suggested answer and vote split |
Open either one in Notepad, VS Code, or anything else. They are plain UTF-8 text.
The links file is written before the slower question pass starts, so if you get bored and close the window you still keep the links.
The .bat files are Windows-only, but the tool itself is not. Run three commands:
git clone https://github.com/swarnava-dutta/Free-Exam-Dumps.git
cd Free-Exam-Dumps
python3 -m venv .venv && .venv/bin/pip install -r requirements.txtThen, to scrape:
.venv/bin/python main.pyEverything from Step 4 onward is identical.
You do not need to know which provider an exam belongs to. Type the code and the tool works it out.
An exact code goes straight through:
Exam code or name: az-104
Exam: Microsoft Azure Administrator
Provider: microsoft
A partial name gives you a numbered list:
Exam code or name: solutions architect
1. [microsoft] Agentic AI Business Solutions Architect
2. [microsoft] Microsoft Azure Solutions Architect
3. [ibm] IBM Cloud Solutions Architect v3
4. [huawei] HCIP-Cloud Service Solutions Architect V3.5
5. [amazon] AWS Certified Solutions Architect - Professional
6. [amazon] AWS Certified Solutions Architect - Associate SAA-C02
7. [amazon] AWS Certified Solutions Architect - Associate SAA-C03
8. [amazon] AWS Certified Solutions Architect - Professional SAP-C02
Pick a number, or press Enter to search again:
Type a number and press Enter. Press Enter on its own to search again instead.
A brand-new exam is fetched from the site automatically. The saved exam list never expires, so an exam added since you built it is not on it. Rather than telling you it does not exist, the tool rechecks examtopics.com and searches again:
Exam code or name: hpe7-a07
'hpe7-a07' is not on the saved list. Checking examtopics.com...
2252 exams indexed.
Exam: HPE Campus Access Mobility Expert
Provider: hp
That recheck happens at most once per run (~25s), so if you simply mistyped, the next attempt is instant:
Exam code or name: saa-c99
'saa-c99' is not on the saved list. Checking examtopics.com...
2252 exams indexed.
No exam matches 'saa-c99'. Try an exam code such as SAA-C03.
Exam code or name: zzznotreal
No exam matches 'zzznotreal'. Try an exam code such as SAA-C03.
Dashes, spaces and capitals do not matter. SAA-C03, saa c03 and saac03 all work.
Measured on a normal home connection:
| Building the exam list, first run only | ~30s, then reused |
Small exam, e.g. github-actions (14 questions) |
~9s |
Medium exam, e.g. hpe7-a07 (18 questions, 96 pages) |
~17s |
Large exam, SAA-C03 (1019 questions, 604 pages) |
169s |
Speed is limited by ExamTopics, not by the tool. Scanning runs at about 8 pages/second and question fetching at about 11 questions/second. Using more parallel workers was measured and does not help — see the notes at the bottom.
One file, .examtopics_index.json, about 160 KB. That is the list of 2252 exams and
which provider each belongs to, saved so the ~190 requests behind it are not repeated on
every run. It has no expiry, and it is refreshed automatically when a search misses.
Nothing else is stored. Every listing page, question page and question count is fetched live on every run, so what you get is always current.
Earlier versions cached the pages too. It was removed: within a single run every page is fetched exactly once, so the cache never sped up the run you were waiting on — it only helped a later run, in exchange for 437 MB of disk and three bugs (a stale question count that cut scans short, unbounded growth, and a sweep that expired the wrong tier). Deleting it made a full SAA-C03 run faster, 192s to 169s, because it no longer writes 250 MB of HTML while you wait.
What that trade gives up, honestly: re-running the same exam is no longer near-instant, a second exam from the same provider no longer reuses the first one's listing pages, and an interrupted run starts over rather than resuming.
Output files are overwritten each run, so move or rename anything you want to keep.
| What you see | What to do |
|---|---|
'python' is not recognized |
Python is not installed, or you missed Add python.exe to PATH. Re-run the Python installer and tick that box. |
[ERROR] Could not create the virtual environment. |
Same cause as above. Install Python 3 from python.org, then re-run install_dependencies.bat. |
[WARN] .venv not found. |
You skipped Step 2. Double-click install_dependencies.bat first. |
| The window flashes open and closes instantly | Do not run the .py file directly. Use run_scraper.bat, which keeps the window open. |
No exam matches '...' |
The tool has already rechecked the site, so the exam genuinely is not there under that name. Type part of the exam name instead and pick from the list. |
is not on the saved list. Checking examtopics.com... |
Normal. The exam was added after you built the list, so it is being rebuilt. ~25s, once per run. |
No discussion links found for this exam. |
Nobody has posted discussions for that exam yet. There is nothing to download. |
[WARN] N pages failed after retries. |
A few requests were refused. The output is still written, just short by those pages. Wait a minute and run it again. |
Fewer questions than the Published: count |
Normal. See "What this cannot do" below. |
It feels stuck on Loading exam index |
It is fetching 188 pages. Give it 30 seconds. It only happens on the first run, or after deleting .examtopics_index.json. |
Be clear-eyed about this before you rely on it.
-
It does not bypass anything. No login, no contributor access, no captcha solving, no paid tier. It reads pages that are already public to anyone with a browser.
-
You only get questions the community has discussed. This is the big one, and it varies enormously by exam. Every run ends with the number stated plainly:
Wrote github-actions dumps.txt (14 questions) Coverage: 14 of 99 published (14%)Measured coverage:
Exam Got Published Coverage Question numbers saa-c031019 1019 100% 1–1019, no gaps dva-c02557 557 100% 1–557, no gaps gh-300103 116 89% 7 topics, 6 gaps hpe7-a0718 62 29% 1–59, 41 gaps github-actions14 99 14% 1–52, 38 gaps Popular exams are complete. Niche ones are sparse, because a question nobody discussed has no public page to read. Nothing can fix that from outside the paywall.
-
Answers are not authoritative.
Suggested answeris ExamTopics' answer and it is sometimes wrong; that is exactly why the community vote is shown next to it. Use both. -
Questions can go stale. Exams get updated. Nothing here knows which questions are still current.
There is one test file with 9 checks. It needs no network and no test framework:
.venv\Scripts\python.exe test_examtopics.py
ok test_block_detection_does_not_eat_real_pages
ok test_exam_index_extraction_and_search
ok test_exam_matching_is_anchored
ok test_multi_answer_question
ok test_page_and_question_counts
ok test_provider_normalization
ok test_question_parsing
ok test_question_parsing_tolerates_missing_parts
ok test_url_rebuild_and_ordering
9 checks passed.
Run this after changing anything. Each check guards a bug that actually happened.
Every decision below came from measuring the live site, not from guessing.
- Questions come from discussion pages, not exam pages.
/exams/<provider>/<exam>/view/is free for page 1 only; page 2 onward is a paywall. Discussion pages are free and carry the question, the choices, the suggested answer and the vote tally. - The suggested answer is the choice ExamTopics tags with
correct-hiddenin its HTML, which is what its paywalled "Show Suggested Answer" button reveals. "Choose three" questions tag several, so all of them are collected. - There is no per-exam discussion listing, so finding one exam's questions means scanning the provider's listing pages. The scan stops early once every published question has been found — SAA-C03 stops at page 480 of 604.
- Exam matching is anchored on
exam<slug>topic. Without the anchor,aws-certified-developer-associatealso matches every...-dva-c02question. - Link text is matched, not just the URL. ExamTopics truncates long URLs before the
-topicanchor, so SAA-C03 matches 0 of its entries by URL and all 1019 by link text. - 24 parallel workers. 16, 32, 64 and 96 were all measured: every one lands on ~6-8 requests/second, so the limit is server-side and extra workers only add sockets.
- Blocked-page detection requires the site nav to be missing. Searching whole pages for phrases like "access denied" silently discarded real AWS IAM questions — 2 of 1019 on SAA-C03.
- Pages are never cached. Only the exam list is saved. Caching pages could not speed up the run you are waiting on, because each page is fetched once anyway — and it cost 437 MB plus a stale-count bug that cut scans short. Removing it made SAA-C03 go 192s to 169s.
- The published question count is fetched live, because it decides when the scan stops early. Reading a stale low count stops the scan short: feeding the scanner an understated 200 on SAA-C03 returned 556 links instead of 1019.
- Failures get a second slow pass. Anything still missing after that is reported as a
[WARN], never silently dropped.
Repeatability was checked rather than assumed. Across saa-c03, dva-c02, gh-300,
hpe7-a07 and github-actions:
- The early stop was compared against a full scan of every page on all five exams. Same links both ways, every time — it has never been observed to lose a question.
- Two independent scans of
saa-c03, both downloading all 604 listing pages fresh, returned identical 1019-link sets. - Two independent cold fetches of all 1154 questions across four exams produced identical results, with zero entries missing question text, choices, a suggested answer or a community answer, and no duplicate topic/question pairs.
One inherent caveat: the provider listing is ordered by last activity, so pages can shift while a scan is in flight. No drift was observed in testing, and duplicates are removed either way, but a very active provider could in principle move an entry across a page boundary mid-scan. The coverage line is your check on that.
main.py terminal app
install_dependencies.bat one-time setup (Windows)
run_scraper.bat launcher (Windows)
requirements.txt requests, tqdm
test_examtopics.py offline checks
examtopics/index.py provider and exam index, exam code lookup
examtopics/scanner.py listing scan and question fetch
examtopics/parsers.py HTML extraction
examtopics/matching.py slugs and exam matching
examtopics/http_client.py retrying GET, thread pool
examtopics/output.py the two output files
examtopics/settings.py tunables
Want it faster or slower? Everything adjustable lives in examtopics/settings.py:
WORKERS, TIMEOUT, RETRIES, CHUNK_PAGES and RETRY_WORKERS.
See LICENSE.