v0.5.6 · Public Beta

Full Reference

Every command, every integration, every benchmark.

Commands Accuracy Integrations MCP Server CI/CD Watchdog GitHub Action
Commands

25 commands. One CLI. Zero dependencies.

Every command works the same way: npx thuban [command] [path]

QUICK START

Run these in order to see what Thuban does:

npx thuban scan . Scan your project
npx thuban fix . Preview what can be fixed
npx thuban fix . --fix Apply all safe fixes
npx thuban dashboard . Open visual HTML report

SCAN & DETECT

thuban scan . Full health scan — score, issues, hallucinations, grade thuban hallucinate . Find phantom APIs, invented methods, fake imports PRO thuban ghosts . Find dead code — functions that exist but nobody calls PRO thuban ai-score . AI risk score per function — which code needs human review PRO thuban clones . Copy-paste drift — find duplicate code that should be shared PRO thuban drift . Detect files that have drifted from their declared purpose PRO thuban deps . Map dependencies, detect circular imports, find orphans thuban diff . Scan only git-changed files — fast CI check thuban debt . Detailed tech debt breakdown with severity scores thuban taint . Trace untrusted data flows through your codebase PRO thuban compare . ../other Side-by-side health comparison of two codebases PRO

FIX & PROTECT

thuban fix . Preview what can be fixed — safe, no changes made PRO thuban fix . --fix Apply all safe fixes automatically PRO thuban fix . --fix --commit Each fix = 1 git commit — revert any with git revert PRO thuban gate --fix Install pre-commit hook — block hallucinated code automatically PRO thuban inject . Inject Mother Code DNA annotations into every file PRO thuban baseline . Snapshot current issues — future scans only flag NEW problems TEAM thuban watch . Live sentinel — monitors your codebase in real-time PRO

REPORT & ANALYSE

thuban dashboard . Generate beautiful HTML dashboard with score rings PRO thuban executive . CTO-ready PDF report — forward it to your board PRO thuban cost . Tech debt cost in $ and hours — ROI in one command PRO thuban passport . Generate codebase identity file — new devs productive in minutes PRO thuban health . Quick health check with summary scores thuban report . Full combined report — scan + deps + drift in one go thuban history . View scan history — track health over time PRO thuban trend . Visualise health trends — improving or degrading? PRO thuban monitor . Scheduled recurring scans with alerts TEAM

OPTIONS

--fix Apply fixes (default is dry-run preview) --unsafe Allow runtime-changing fixes (circular deps, deprecated API swaps) --commit Auto-commit each fix as a separate git commit for easy rollback --json Output as JSON for CI/CD pipelines --baseline Only report NEW issues vs saved baseline --verbose Detailed output with extra context

We publish our accuracy because transparency builds trust

Most scanners advertise detection rates and stop there. We show you precision against a labelled corpus you can inspect yourself.

1.0
Precision

Zero false positives against our labelled test corpus. When Thuban flags something, it's real.

Precision 1.0 — zero false positives. When Thuban flags something, it is real. This isn’t a marketing number — it’s measured against a real, versioned test corpus of known-good and known-bad code across 10 languages, and it moves with every release. High precision means you can trust every finding without triage fatigue.

Methodology: precision is computed by running the Scanner and Crucible engines against a versioned, labelled corpus of known-good and known-bad code samples spanning all 10 supported languages. Precision is the share of flagged issues that are real (true positives / all flagged). The corpus and scoring scripts are public so you can reproduce the numbers yourself.

View the public benchmark repo →

Integrations

Works where you work

CLI, IDE, CI/CD, and AI agents. Thuban fits into your existing workflow without friction.

💻

VS Code Extension

Real-time scanning on save. Issues appear as diagnostics in the Problems panel. Quick-fix suggestions via Code Actions. Trust score in the status bar.

code --install-extension thuban-0.5.6.vsix Download VSIX →
🤖

MCP Server

Let AI agents call Thuban directly. 7 tools including scan, fix, hallucinate, ghosts, clones, taint, and health. Works with Claude, Copilot, Cursor.

npx thuban-mcp View source →

GitHub Action

Automatic PR scanning. Comments with trust score delta, new issues, and inline annotations. Optional auto-fix PRs. Configurable quality gates.

uses: SilverwingsBenefitsGit/thuban-action@v1 View source →

The verification layer for AI coding agents

Nobody owns this category yet. Thuban’s MCP server lets Claude, Cursor, Copilot, and any AI coding agent verify code as they generate it — not after. The agent writes code, Thuban checks it, the agent fixes it before a human ever sees it.

🤖
AI Agent writes code
Claude, Cursor, Copilot
Thuban MCP verifies
7 tools, instant results
Issues found?
Agent fixes them
Clean code delivered
Human reviews verified code

7 tools your AI agent can call

thuban_scan

Full codebase scan — health score, issues, grades

thuban_fix

Auto-fix issues with safe, reversible git commits

thuban_hallucinate

Detect fake APIs, phantom imports, invented packages

thuban_taint

Track user input from source to dangerous sink

thuban_ghosts

Find dead code, orphan files, unused functions

thuban_clones

Spot duplicated logic AI copy-pasted across files

thuban_health

Quick health check — score, grade, issue count

🎯 Why this matters

AI agents generate thousands of lines of code per day. Without verification, hallucinated APIs ship to production, phantom packages get imported, and plausible-but-wrong logic passes code review because it looks right. The MCP server makes verification automatic — the AI checks its own work before you ever see it.

🚀 The category is wide open

SonarQube and Snyk were built for human bugs. Nobody owns “the verification layer for AI-generated code” yet. Thuban is purpose-built for the vibe-coding era — where the code is written by machines and the humans need to know it’s safe. This isn’t a feature. It’s infrastructure.

Set up in 30 seconds

# Install globally
npm install -g thuban-mcp
# Add to your AI agent config
{
"mcpServers": {
"thuban": {
"command": "thuban-mcp"
}
}
}
View on GitHub → npm →

A digital watchdog for every line of code your team writes

Thuban scans every single push against both engines — the Scanner and the Crucible — automatically, checking for known patterns. It's an automated first pass, not a substitute for human code review.

💻
Developer pushes
Any branch, any language
Thuban Scanner
69 rules, security + quality
Crucible Engine
515 attack patterns
Verified & scored
Trust score on every PR
🔒

Nothing ships unverified

Every PR gets a trust score. Block merges below your threshold. SQL injection in line 42? Blocked before it reaches staging. AI hallucinated an API? Caught at push time, not production.

👥

Scale without fear

10 devs or 500. Junior devs, contractors, AI agents — they all push code. Thuban checks every single one equally. No bias, no fatigue, no Friday afternoon shortcuts. The watchdog never sleeps.

📈

Track quality over time

Every scan builds a trust history. Watch your codebase health improve sprint by sprint. Show the board a graph that goes up and to the right — not just revenue, but code quality.

💬

For CTOs managing teams of any size

You can't manually review every PR. You can't catch every hallucinated API an AI copilot generates. You can't be sure that contractor on the other side of the world isn't introducing SQL injection at 2am. But Thuban can.

Wire it into your CI pipeline in 30 seconds. From that moment, every push against every branch runs through 69 security and quality rules, then gets stress-tested against 515 attack patterns. Your team gets instant feedback. You get a dashboard showing trust scores across every project. No meetings. No overhead. Just verified code, every time.

GitHub Action GitLab CI Bitbucket Pipelines Jenkins Any CI/CD
# Add to any CI pipeline — one line
npx thuban scan . --ci --fail-below 75

Drop one YAML file into your repo

Runs on every PR. Blocks merges on critical findings. Posts the health score straight into the PR conversation.

# .github/workflows/thuban.yml name: Thuban Code Scan on: [pull_request] permissions: contents: read pull-requests: write jobs: scan: runs-on: ubuntu-latest steps: - uses: actions/checkout@v4 - uses: thuban-dev/thuban-scan@v1 with: fail-on: high self-verify: true advisor: true

Runs on every pull request — no separate install step, no Docker, no API keys.

🚫

Blocks merges when findings breach your fail-on severity threshold.

💬

Optional Slack notifications for critical findings via --slack-webhook.

Full CI/CD Integration docs →

Protect your codebase in the next 60 seconds.

No install, no config files, no signup required.

$ npx thuban scan .

Click to copy

Thuban Help
Ask a question or pick a topic below.