CodeTwin ML • Local-first software intelligence · v1.0.0

Understand your code. Fix with proof.

Build a persistent Software Digital Twin of your repository, trace dependencies, review security and quality findings, and approve verified repairs. Your source stays on your device.

✓ Local-first✓ Source-linked evidence✓ Optional local AI
Windows x64 stable releaseNo bundled AI model
CodeTwin ML — project analysis
Local workspace
auth.ts
Impact graph
Findings
01export async function authorize(request) {
02 const actor = await session.user()
03 if (!actor) return deny()
04 return policy.evaluate(actor)
05}
06// Findings require supporting evidence
ANALYSIS · project index completed
Symbols · dependencies · change impact
EVIDENCE · uncertain checks are not passes
Illustrative interface — not a live scanLocal-first · Evidence-aware
Local-firstYour code remains on-device
Evidence-drivenOnly evidence-backed conclusions
User-approved repairsReview every proposed change
Windows x64 releaseOfficial organization releases
Code intelligence meets security

A complete picture of the project.

CodeTwin connects indexed files, dependency graphs, findings, test evidence and approved repairs inside one software digital twin.

Code intelligence

Software Digital Twin

Indexes files, symbols, dependencies and references using Tree-sitter, with SHA-256-backed incremental updates and impact analysis.

Tree-sitterSHA-256Impact
Security

Static security analysis

Highlights credential-like literals (redacted), dangerous execution patterns, weak hashes and cross-language risk signals without running the project.

StaticRedactedEvidence
Security

Dependency vulnerability checks

Reviews dependency risks using OSV data and reports evidence and uncertainty rather than turning incomplete checks into passes.

OSVDependencies
Quality

QA and test evidence

Discovers test configuration separately from execution; Windows can run explicitly selected tests in a restricted sandbox.

DiscoveryWindowsSandbox
Security

Authorized web security

Guided scope-limited assessments for sites you own or are permitted to test, with operator approval and source-linked findings.

Approved scopeWeb
Remediation

Fix & Verify

Review an exact patch, approve it, apply with backups, re-index and retest. A fix is not marked verified without successful retesting.

ApprovalRollbackRetest
How it works

From source to verified change.

Uncertain results are not silently reported as passes. A generated patch is never automatically a verified fix.

STEP 01

Add a repository

Select a local project. CodeTwin indexes supported source files, maps symbols and records dependency evidence without running the project.

STEP 02

Analyze the digital twin

Inspect quality, security, database, runtime and QA evidence. Review what is confirmed, missing, skipped or inconclusive.

STEP 03

Approve, fix and verify

Review bounded patches and exact diffs. Approve changes, retain backups and require post-change retests before calling a finding fixed.

Privacy and control

Local by design. Evidence before action.

Source indexing and analysis run locally. Active website assessments require authorization and bounded scope; local AI is optional and needs a separately installed compatible model.

✓Passive analysis does not execute project code
✓Secret-like values are redacted in findings
✓Patch review, approval, backup and retesting
✓Evidence distinguishes confirmed and inconclusive outcomes
SHA
Source integritySHA-256 file snapshots
Controlled
QA
Test evidenceDiscovery ≠ execution
Controlled
WEB
Scoped testingPermission required
Controlled
AI
On-device modelsOptional ONNX / GGUF
Controlled
Common questions

Know what the software actually does.

Does CodeTwin upload my source code?

Local indexing, static analysis and on-device ML work on your machine. Optional online lookup features, such as a dependency database check, may require network access; inspect each enabled integration before use.

Which operating systems have official installers?

The published v1.0.0 binary release contains Windows x64 EXE and MSI installers. Linux .sh and macOS .command files are launch scripts, not native application installers.

Is an AI model included?

No. On-device AI features require a separately installed, integrity-checked compatible ONNX or GGUF model.

Can CodeTwin change files automatically?

Repair workflows require explicit review and approval of the exact patch before applying it. Source-hash checks, backups and follow-up verification provide additional safeguards.

Can I scan any website?

No. Active website testing is for properties you own or have explicit permission to test, with a bounded scope and approval steps.

Does detecting a test mean it passed?

No. Test discovery does not mean execution. A pass/fail verdict requires a completed and interpretable test run; inconclusive outcomes remain inconclusive.

Official stable release

Download CodeTwin ML.

Get installer files only from the OpenMind AI Technologies Limited GitHub release. Current stable: v1.0.0.

Windows x64 installer

Windows EXE setup

Recommended for a standard desktop installation.

CodeTwin-ML_1.0.0_x64-setup.exe · 182.5 MBDownload EXE ↗
Windows x64 installer

Windows MSI package

Alternative installer for managed Windows environments.

CodeTwin-ML_1.0.0_x64_en-US.msi · 254.6 MBDownload MSI ↗

Verify installer integrity using the official SHA256SUMS.txt file. Compare its hash to your downloaded installer.