Which AI coding assistants does aiverify work with?

aiverify works with code generated by any AI assistant — Claude, GitHub Copilot, Cursor, ChatGPT, Tabnine, Replit AI, Amazon CodeWhisperer, and more. It doesn't matter which AI wrote the code; aiverify finds the telltale patterns of AI-generated code.

How is aiverify different from pylint or flake8?

Traditional linters like pylint and flake8 find style and syntax issues. aiverify specifically targets AI hallucination patterns: invented function names, deprecated APIs that AIs frequently use, swapped argument orders, hardcoded API keys, infinite loops, and commented-out dead code. These are patterns unique to AI-generated code.

How do I install aiverify?

Install aiverify with pip: 'pip install aiverify-cli'. Requires Python 3.11+. Then run 'aiverify your-code.py' on any Python, JavaScript, or TypeScript file.

Yes! aiverify is completely free and open source under the MIT license. It's built by a solo founder and supported by GitHub Sponsors.

Can I use aiverify in CI/CD?

Yes. aiverify supports JSON output with the --json flag and returns exit code 0 (pass) or 1 (issues found). It integrates natively with GitHub Actions, GitLab CI, pre-commit hooks, and any CI/CD pipeline.

How many rules does aiverify have?

aiverify has 15 detection rules: 5 for Python (unused imports, hallucinated functions, deprecated APIs, wrong argument order, type mismatches), 5 for JavaScript/TypeScript (same categories plus null/undefined detection), and 5 generic rules for all languages (hardcoded credentials, infinite loops, unused variables, commented dead code, magic numbers).

🔍 Open Source · MIT · Zero Config

Your AI coder is hallucinating.
Here's the proof.

Claude, Copilot, Cursor, ChatGPT — they all invent functions, leak API keys, and use deprecated APIs. aiverify catches it in 3 seconds. pip install aiverify-cli

⭐ ...

GitHub Stars

👁️ ...

Repo Visitors

🔍 15

Detection Rules

🌐 3

Languages Supported

Install Now Star on GitHub ❤️ Sponsor

aiverify — examples/bad_python.py

$ aiverify examples/bad_python.py

aiverify — AI Code Verification
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
✗ GEN001 Hardcoded OpenAI API key (sk-...)
→ examples/bad_python.py:15:12
⚠ PY001 Unused import: 'os'
→ examples/bad_python.py:6:1
⚠ PY002 Function 'calculate_metricz' is called but not defined
→ examples/bad_python.py:19:15
⚠ PY003 Deprecated import 'pkg_resources'
→ examples/bad_python.py:11:1
⚠ PY005 Type mismatch: string + integer
→ examples/bad_python.py:29:9
ℹ GEN002 Potential infinite loop: 'while True' without 'break'
→ examples/bad_python.py:34:1
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
20 issues found | 2 errors | 17 warnings | 1 info

$ echo "AI hallucination caught. Code rejected by CI."

Why aiverify?

15 tree-sitter-powered rules that catch what traditional linters miss. Built specifically for AI-generated code.

🧠

Hallucination Detection

Catches invented function names, phantom imports, and API calls that don't exist. AI models hallucinate 15-30% of code — aiverify proves it.

🔑

Credential Leak Prevention

Detects hardcoded API keys (sk-...), passwords, AWS secrets, database URLs, and tokens before they ship to production.

🌳

Tree-Sitter AST Parsing

Not regex. Full Abstract Syntax Tree parsing — the same engine used by VS Code. Accurate, false-positive-free analysis.

⏱️

3-Second Scans

Zero configuration. Point at any Python, JavaScript, or TypeScript file and get results instantly. Perfect for pre-commit hooks.

🔌

CI/CD Native

JSON output with --json, exit code 0/1, and GitHub Actions integration. Block AI-hallucinated code from shipping automatically.

🆓

Free & Open Source

MIT licensed. Zero cost. Zero API keys. Built by a solo founder who got tired of debugging AI-generated code in production.

15 Detection Rules

Every rule uses tree-sitter AST queries — not fragile regex — for accurate, reliable detection with minimal false positives.

🐍 Python

PY001unused-import — Imported module never used anywhere in the file

PY002hallucinated-func — Function called but not defined, imported, or built-in

PY003deprecated-api — pkg_resources, distutils, imp, inspect.getargspec, logging.warn

PY004wrong-arg-order — Duplicate keyword args, suspicious argument patterns

PY005type-mismatch — str + int, len() compared to string, integer dict keys

🌐 JavaScript / TypeScript

JS001unused-import — Imported symbol never referenced

JS002hallucinated-func — Function call with no definition or import

JS003deprecated-api — substr(), createClass, console.exception, Array.observe

JS004wrong-arg-order — Callback params swapped: (data, error) instead of (error, data)

JS005null-undefined — Property access on variables without null checks

🔧 Generic (All Languages)

GEN001hardcoded-creds — API keys, passwords, tokens, database connection strings

GEN002infinite-loop-risk — while True without break or return in reachable path

GEN003unused-variable — Variable assigned but never read

GEN004commented-deadcode — Large blocks of commented-out code (failed AI generation)

GEN005magic-number — Same numeric literal 3+ times with no named constant

How It Works

Tree-sitter parses your code into an AST. Rules inspect the AST for patterns unique to AI-generated code.

Parse

aiverify reads your file and builds a full Abstract Syntax Tree using tree-sitter — the same parser used by VS Code, GitHub, and modern IDEs.

Analyze

15 rules run against the AST, each looking for specific patterns: invented functions, deprecated APIs, type mismatches, hardcoded credentials, and more.

Report

Color-coded results with file paths and line numbers. JSON output for CI/CD. Exit code 1 blocks bad code from shipping.

aiverify vs Alternatives

Existing linters find style issues. aiverify finds the issues that prove your code was written by an AI.

Feature	aiverify	pylint	flake8	eslint	semgrep
AI hallucination rules	✅ 15 rules	❌	❌	❌	❌
Invented function detection	✅	❌	❌	❌	❌
Hardcoded credentials	✅	❌	❌	❌	✅
Infinite loop detection	✅	❌	❌	❌	❌
Commented dead code	✅	❌	❌	❌	❌
Magic number detection	✅	❌	❌	❌	❌
Tree-sitter AST (not regex)	✅	❌	❌	❌	✅
Zero config, 3s scans	✅	❌	❌	❌	❌
Python + JS + TS	✅	🐍 only	🐍 only	🌐 only	✅

FAQ

Everything you need to know about aiverify.

aiverify is a free, open-source CLI tool that detects AI hallucinations in code. It scans Python, JavaScript, and TypeScript files using tree-sitter AST parsing and catches issues like invented functions, deprecated APIs, hardcoded credentials, type mismatches, and infinite loops.

aiverify works with code generated by any AI assistant — Claude (Sonnet, Opus), GitHub Copilot, Cursor, ChatGPT (GPT-4, GPT-4o), Tabnine, Replit AI, Amazon CodeWhisperer, Codex, and more. It doesn't matter which AI wrote the code — aiverify finds the telltale patterns of AI generation.

Traditional linters find style and syntax issues — unused variables, missing semicolons, formatting problems. aiverify specifically targets AI hallucination patterns: invented function names that don't exist, deprecated APIs that AIs frequently suggest, swapped argument orders, hardcoded credentials, infinite loops, and blocks of commented-out dead code. These are patterns unique to AI-generated code that traditional linters completely miss.

Install: pip install aiverify-cli (PyPI package name, requires Python 3.11+). The CLI command is aiverify. Then run: aiverify your-code.py. To scan a directory: aiverify src/. For CI/CD: aiverify src/ --json. To filter specific rules: aiverify app.py --rules PY002 GEN001.

Yes, completely free and open source under the MIT license. This is a solo founder project — it's supported by GitHub Sponsors. If aiverify saves you time or prevents a production incident, please consider sponsoring or starring the repo.

No. aiverify uses tree-sitter, the same parsing engine used by VS Code, GitHub, and modern IDEs. Tree-sitter builds a full Abstract Syntax Tree (AST) for accurate, reliable analysis with minimal false positives. Regex-based parsing is fragile and breaks on formatting changes — aiverify doesn't use it.

Yes. aiverify supports --json flag for machine-readable output and returns exit code 0 (clean) or 1 (issues found). It integrates natively with GitHub Actions, GitLab CI, pre-commit hooks, and any CI/CD system. See the GitHub README for CI configuration examples.

Currently: Python (.py), JavaScript (.js, .jsx), and TypeScript (.ts, .tsx). Go, Rust, Java, and Ruby support are on the roadmap — the more stars and community interest, the faster they ship.

Install in 3 Seconds

One command. No dependencies beyond Python 3.11+. Works on Linux, macOS, and Windows.

pip install aiverify-cli

Then run: aiverify your-code.py

or scan a whole project: aiverify src/

📎 Embed badge in your README

[![aiverify](https://img.shields.io/badge/verified%20with-aiverify-58a6ff?style=flat-square)](https://fmathenomad.github.io/aiverify/)

Your AI coder is hallucinating.Here's the proof.