NEXADiag · NexaVerify
User Manual

NexaVerify — How it works

A complete walkthrough of NexaVerify: provider setup, scan modes, the consensus engine, the HTML and JSON reports, free tier limits, and the Pro license. Reflects exactly what's in the version available on Gumroad today.

Start Providers Scan Report Health Score License Free Tier FAQ
Current public version
Multi-LLM Code Audit

Built for indie devs who don't trust a single AI to review their code. 4 frontier LLMs vote in parallel. Only signals that hold up survive.

01 · Getting Started

First launch — the Control Center

When you open NexaVerify for the first time, you land directly on the Control Center. No long tutorial. No wizard. A clean dark dashboard with one job: get you from "app open" to "first scan" in under 60 seconds.

1
Notice the API key indicator
If no API key is configured, a discreet banner at the top invites you to open Settings. You can't run a scan without at least one active provider.
2
Open Settings
Click the gear icon. Paste your API key for at least one provider. For best consensus quality, configure all four (see next section).
3
Pick a project folder
Browse to the folder you want to audit. Real codebase, not an empty test folder.
4
Choose a mode and click Launch Radar
Quick for a fast surface check. Balanced for the daily default. Deep Analysis when stakes are high.
Tip: start with Balanced. Move to Deep Analysis only for pre-delivery audits or when something in Balanced needs cross-checking.
02 · Scan modes

Three modes, three intents

Each mode controls how many providers are called and how aggressively the consensus engine cross-checks results.

Quick Scan

Surface analysis. Catches obvious syntax issues and clear logic errors. Use it as a 2-minute sanity pass before continuing.

Balanced Scan ★

Default mode. Cross-references results across two providers. Best tradeoff between speed, signal quality, and API cost.

Deep Analysis

Full consensus engine. All four configured providers are queried in parallel. Use it before a client delivery, on critical code, or when Balanced surfaces something that needs confirmation.

03 · The report

HTML for humans, JSON for machines

Every scan generates two outputs at the same time. You decide which one to use depending on the task.

  • HTML report — interactive, styled in a dark theme, with a Health Score, severity colors, confidence bars, and the list of validated issues. Best for client handoffs and human reading.
  • JSON export — raw data for archiving, automation, CI/CD pipelines, or comparing multiple runs over time.

Each issue in the report includes: file and line number, severity (CRITICAL / HIGH / MEDIUM / LOW), confidence percentage, and which providers detected it.

Reading tip: a finding flagged by 3 of 4 providers at HIGH severity is almost always a real bug. A solo finding at LOW severity is a hint, not a verdict.
04 · Health Score

How the score is calculated

The Health Score starts at 100 and is reduced based on validated issues, weighted by severity and confidence. Solo detections (1 provider only) reduce the score less than multi-LLM agreements.

Issue typePenalty (full confidence)
Critical issue−15 pts
High issue−8 pts
Medium issue−3 pts
Low issue−1 pt
Partial confidence (1 LLM only)Reduced proportionally

A "Why this score?" button is available in every HTML report. It explains which findings contributed to the deduction. The score is a guide, not an absolute verdict — the issue list is what matters.

05 · Licensing

One-time purchase, lifetime access

NexaVerify is sold as a one-time license, not a subscription. Buy once on Gumroad, use forever. Your license unlocks the Pro tier on the machine where you activate it.

1
Purchase on Gumroad
After payment, Gumroad sends you the download link and your license key by email.
2
Paste the key in NexaVerify
Open the app, go to Settings → License, paste the key. The Pro tier unlocks immediately.
3
License is machine-bound
Each license activates on one PC. If you change machine, contact nexadiag@gmail.com with proof of purchase for a reset.
Free updates: when a new version of NexaVerify ships, existing buyers receive it automatically through Gumroad. No extra payment, no expiration.
Providers

4 LLMs, your own keys

NexaVerify uses your own API keys. Code goes directly from your machine to the provider's endpoint. No NEXADiag server in between, no proxy, no telemetry.

OpenAI
GPT-4o / Turbo
Strong on syntax and architectural patterns.
Anthropic
Claude 3.5 Sonnet
Strong on logic, security, and edge cases.
Google
Gemini 1.5 Pro
Strong on context and large file analysis.
Groq
Llama 3
Free tier available. Fast inference, useful for the consensus speed.
Best practice: configure all 4. The consensus engine becomes meaningfully stronger with 4 different model architectures voting. Two is the minimum to absorb provider failures (quota, rate limit, downtime).
Free Tier

Try before you buy

The free version is for evaluation, not production. It runs the same engine but with hard limits.

  • 3 analyses per day — quota resets at local midnight based on your PC date.
  • 10 files maximum per scan — to keep free-tier API calls reasonable on shared free quotas.
  • No credit card required. Just download from Gumroad and run.
  • Pro license removes both limits.
Privacy

What stays local, what goes out

  • Your code is sent only to the LLM providers you configured, using their official APIs.
  • Your API keys are stored locally on your machine. NEXADiag never sees them.
  • Reports are generated locally and saved on your machine. Nothing is uploaded.
  • No telemetry, no analytics, no tracking pixels.
Internet is required for the actual analysis (the LLM API calls need connectivity). Local-first means no NEXADiag server between you and the providers — not air-gapped.
FAQ

Common questions

Why use 4 LLMs instead of one? ▼
Single-model reviews hallucinate and miss things. Each model has different blind spots: GPT can miss architectural issues, Claude can miss certain security patterns, Gemini can miss subtle logic bugs, Groq is fast but less precise. Running them in parallel and surfacing only what survives consensus reduces false positives drastically.
Do I really need 4 keys to start? ▼
No. NexaVerify runs with just one provider configured. But the consensus engine becomes meaningful only with 2 or more. Groq's free tier is the easiest no-cost entry point if you don't already have other API keys.
What happens if one provider fails during a scan? ▼
The scan continues with the providers that responded. The report is still generated, marked with "partial confidence" so you know the consensus was based on fewer signals.
Where is the report saved? ▼
In your local NexaVerify folder, alongside the JSON. The HTML opens in your default browser automatically after each scan.
Linux or macOS? ▼
Windows 10/11 64-bit only for the current version. Wine on Linux is experimental. macOS is not yet supported.
Will this tool keep evolving? ▼
Yes. NexaVerify is in active development. Buyers receive future versions automatically through Gumroad. The roadmap is shaped by real user feedback — every email at nexadiag@gmail.com gets a reply.
NexaVerify by NEXADiag · User Manual · ← Back to landing Get Lifetime Access Contact