New: both AI detection models just got a major upgrade. They now catch ChatGPT, Claude, and the newest AI hiding in student code. See what's new
Platform Documentation

How Codequiry works

From upload to verdict in a few minutes. Run code through peer comparison, an extensive web search, and AI detection, then read every match side by side.

50+
Programming Languages
3
Detection Engines
Live
Real Time Progress
200
Submissions per Check
The Workflow

The life of a check

Every check moves through the same five stages. You watch each one happen in real time from your dashboard.

Create

Add a folder, then run a new check. Pick the language and detection type.

New

Upload

Add submissions as ZIP files. The check is now ready to start.

Ready for checking

Queue

Hit Start. Your check is handed to an analysis server.

Waiting for server

Analyze

Engines parse and compare every file. Logs stream live.

Checking in progress

Review

Results are ready. Open the gauge, composition, and matches.

Completed

The status label on each check in your dashboard tracks exactly where it is, from New to Completed. If a submission cannot be parsed it is flagged so you can fix the upload and run again.

Core Concepts

Folders, Checks, Submissions

The whole platform is built on three simple objects that nest inside each other.

Folders

Top level containers that keep your work organized. Create one per class, term, or project.

  • Unlimited checks inside each folder
  • Turn on the Student Portal per folder
  • Sort and search across every check

Checks

A check holds the submissions for one assignment and runs the analysis. You set the language and detection type when you create it.

  • Group Similarity, or Web plus Group
  • AI Detection runs alongside the check
  • Live status from New to Completed

Submissions

One ZIP per student. Add them yourself, sync from the Desktop App, or collect them through the Student Portal.

  • ZIP files, up to 10MB each
  • Up to 200 submissions per check
  • View, download, or delete anytime
Detection Engines

Three ways code gets caught

Pick a similarity type when you run a check. AI Detection runs alongside it. Together they cover copying between students, copying from the internet, and AI generated code.

Peer Comparison

Group Similarity Check

Compares every submission against each other, with no web search. The fast, focused option for catching collaboration inside a class or competition.

All against all comparison
Compares up to 500 submissions
Needs 2 or more submissions
Usually done in a few minutes
Web Sources

Web & Group Similarity Check

Everything Group Similarity does, plus an extensive search across public code and the Codequiry Database. Use it when external copying is a concern.

Billions of lines of public code
Links back to each original source
Includes peer comparison too
GitHubGitLabStack OverflowCheggSoftware HeritageCodequiry Database
AI Detection

AI Generated Code

Flags code that reads as machine written. Two analyses run together: a classifier that scores AI versus human, and a style analyzer that looks at how the code is written.

Per file AI and human probability
Style, entropy, and comment signals
Detailed report you can export

When you run a check you choose one of the two similarity types in the create dialog. Matches are tagged by where they came from, Peer, GitHub, Stack Overflow, Web, or the Codequiry Database, so you always know the source.

Real Time

Watch the analysis run

You do not stare at a spinner. While a check runs, Codequiry streams the engine logs and live progress straight to your screen.

Nothing happens in a black box

Every check reports what it is doing as it does it, so a long web search never feels stuck.

Live engine logs

A streaming feed shows each file being parsed and compared in order.

Progress and machine load

A progress bar plus live CPU and memory readouts tell you how far along the run is.

Results when ready

The check flips to Completed the moment analysis finishes. No need to keep the tab open.

check #4821 live log
10:42:01 parsing 38 submissions
10:42:04 ok tokenized main.py, utils.py, model.py
10:42:09 peer comparing submissions all against all
10:42:15 web searching public sources and database
10:42:22 match sub 14 vs sub 27 at 92 percent
10:42:25 ai scoring files for generated code
72 percent completecpu 41 percent · mem 2.1 GB
Reading Results

One score, then the whole story

Results open on an Estimated Unoriginality gauge and a composition breakdown, then let you drill into every match file by file.

High Risk
68%
Unoriginality
Estimated Composition
Peer 55%
Database 25%
Web 20%
Top10 All63 Peer Matches41 Web Matches22
model.py vs Submission 27lines 14 to 96
Peer 92%
utils.py vs github.com/acme/helperslines 1 to 48
GitHub 74%
parser.py vs Codequiry Databaselines 22 to 70
Database 61%

Click any match in the Match Explorer to open a side by side comparison with the similar lines highlighted.

PDF CSV JSON XML
0 to 20 percent

Typically original. Shared imports or boilerplate may still appear.

21 to 50 percent

Review. Often shared starter code or common solutions.

51 to 80 percent

High. Significant overlap. Investigate and document.

81 to 100 percent

Critical. Extensive copying is likely. Open the comparison.

A percentage is a starting point, not a verdict. Always open the matched code in context. Reused starter code and common library calls can read high without being copying.

AI Detection

Two reads on AI generated code

For each file, one model scores how likely the code is AI written, while a second looks at the style and structure that machine written code tends to share.

AI Classification

A probability that the file was machine generated, with a human and AI split.

AI probability82%
Human probability18%

Each file is labelled with a classification so you can sort a whole submission at a glance.

Perfection & Style

Signals in how the code is written, not just what it does.

4.2
Entropy
31%
Comment density
Uniform naming No dead code Textbook structure

An AI score is a signal, not a verdict. Very clean code, short files, and provided starter code can read high. Use it to decide what to look at more closely, not to accuse.

ChatGPT Claude Gemini Copilot and more
Getting Code In

Four ways to collect submissions

Upload it yourself, sync a folder, let students send it directly, or automate the whole thing.

Web Upload

Drop ZIP files onto your check or click to browse. One ZIP per student, up to 10MB each, and up to 200 per check. The analysis server unpacks each archive for you.

Desktop App

Point the app at a local folder and submissions upload as they land. Handy for LMS exports and shared drives. Available for Windows, Mac, and Linux.

Student Portal

No account

Turn on the portal for a folder and share one link. Students enter their name and email, optionally a student ID or section, and upload a ZIP. You can sync due dates from your LMS and brand the page with your logo and color.

REST API

Create checks, upload code, start runs, and pull results straight from your own systems. A single quick check call can create, upload, and start in one request.

Automate it with the API

Everything you can do in the dashboard, you can do from code. Authenticate with an API key and drive the full check lifecycle programmatically.

  • Create checks and pick the detection type
  • Upload single ZIPs or batches
  • Start a run and poll live status
  • Pull overview, matches, and AI results
Read the API docs
quick-check.sh
# Create, upload, and start a check in one call
curl https://codequiry.com/api/v1/check/quick \
  -H "apikey: YOUR_API_KEY" \
  -F "name=Assignment 4" \
  -F "language=python" \
  -F "[email protected]"

# Then poll for progress
curl https://codequiry.com/api/v1/check/status \
  -H "apikey: YOUR_API_KEY" \
  -F "check_id=4821"
Languages

Built for real codebases

Codequiry parses 50+ programming languages and detects each file type automatically.

Python Java JavaScript C++ C C# PHP TypeScript Swift Kotlin Go Ruby Rust SQL and many more
FAQ

Frequently asked questions

Quick answers to the things people ask most.

Navigate to your course dashboard, select the assignment, and click on any submission to view its detailed plagiarism report.
Codequiry supports most programming languages including Python, Java, C++, JavaScript, PHP, and many more. We also support plain text and PDF files.
Most scans complete within 2-5 minutes depending on file size and the number of submissions. Large batches may take longer.
Yes! You can download individual reports as PDF or export batch results as CSV from the assignment page.
Codequiry uses advanced algorithms to compare code structure, logic patterns, and syntax beyond simple text matching.
Essential includes unlimited scans, priority support, API access, advanced reporting, and team collaboration features.
Go to Course Settings > Students, and either invite them via email or share the course enrollment code.
Yes! We support Canvas, Moodle, Blackboard and other LMS platforms. Contact support for integration assistance.
Upload one or more ZIP files containing your source code by dragging and dropping them into the upload area or clicking to browse. Each file can be up to 10MB.
Only ZIP archives are accepted. Your ZIP files can contain source code in any popular programming language including Java, Python, C/C++, JavaScript, C#, PHP, Ruby, Go, Rust, and more.
If your project exceeds 10MB, split it into multiple ZIP files or create separate checks. To reduce file size, exclude build artifacts, dependency folders like node_modules, and version control directories like .git.
Our system compares submissions against each other (peer comparison), our database of past submissions, and optionally against online sources. We use advanced algorithms that detect similarities even when code has been renamed or restructured.
The similarity percentage indicates how much of the submission matches other sources. Higher percentages suggest more overlap. However, some similarity is normal (e.g., common code patterns), so always review the highlighted matches.
Your code is encrypted and secured. We do not store or match your uploaded code in the Codequiry database. Our checks compare against external sources only. Professional plans have additional privacy settings.
Our AI written code detection achieves 80-90%+ accuracy when detecting AI-generated code through multi-layered neural networks trained on millions of code samples. We prioritize false positive reduction, preferring to miss AI-written code rather than incorrectly flag human-written code. When we flag code as AI-generated, it's based on concrete indicators and distinctive patterns that separate machine-generated code from human coding styles.
We can detect AI written code from ChatGPT (GPT-4o, GPT-5, GPT-5.1), GitHub Copilot, Claude (3.5, 4, Opus), Grok, Google Gemini, Cursor AI, Amazon CodeWhisperer, Meta Llama, and all major AI coding assistants. Our General AI Detection model catches any AI-generated code patterns.
We support 65+ programming languages including Python, Java, JavaScript, C++, PHP, Ruby, Go, Rust, Swift, and more.
While no AI detection system is 100% foolproof, our multi-layer fingerprint verification makes it extremely difficult to bypass. We use proprietary pattern recognition trained on LLM writing signatures that go beyond simple text analysis. Students who heavily modify AI-generated code may reduce detection confidence, but distinctive AI patterns often remain. We continuously update our models to detect new evasion techniques.
We prioritize false positive reduction, preferring to miss AI-written code rather than incorrectly flag human-written code. When we flag code as AI-generated, it's based on concrete indicators and distinctive patterns that separate machine-generated code from human coding styles.
AI detection typically takes 10-30 seconds per submission, depending on the code size and complexity. Results are generated in real-time and displayed alongside your plagiarism detection results in a unified report.
Code submissions are processed securely and can be stored in your account for your records and comparison purposes. We never share your code with third parties or use it to train external AI models. You maintain full control over your data and can delete submissions at any time.

Ready to run your first check?

Create a folder, add some code, and read your results in minutes. Trusted by educators to keep programming courses honest.

Go to Dashboard