Veylaro
Sign in Download
Local coding, governed by execution

Your code.
Your machine.
Your AI.

Veylaro is a local coding environment built to read scoped projects, propose focused edits and verify them on your machine. Your code is not sent to a hosted inference service.

Download free →See benchmarks
macOS WindowsFirst month unlimited · No credit card · 2-minute account
100% on-device No usage limits Works offline Agentic coding
Data to cloud0 bytes. Ever.
Usage limits∞ Unlimited
LatencyZero network
Our mission

We believe everyone should own powerful AI.

Not rent it. Not queue for it. Not hand your work to someone else's servers. Intelligence that lives on your machine, answers to you, and gets smarter every night — that's the whole company.

✦ Yours — it lives on your hardware✦ Local by default, with optional online services✦ Verified local precedents stay on your device
Agentic, end to end

An engineer in your terminal.
Not a chatbot in a tab.

Give Veylaro a task in plain English. It plans, edits multiple files, runs your tests, reads the failures and iterates against real evidence. Model inference and project work stay local; live search and account services are optional.

  • Multi-file edits with full project context
  • Runs shell commands, tests and builds
  • Explains its reasoning as it works
  • Git-native: reviewable diffs and clean commits
Explore all features
Local by design

Not connected to the cloud.
Connected to your machine.

VeylaroCode-1.0.0-arm64-mac.zip128 MiB
App download · model weights download separately on first launch
◆ indexing ~/projects/app · on-device
├─ src/ 412 files ✓
├─ server/ 187 files ✓
├─ packages/ui/ 96 files ✓
└─ tests/ 230 files ✓
925 files mapped · 0 uploaded
// auth/session.ts
- const token = md5(user.id)
+ const token = await signJWT(user, expiry)
+ await rotateRefreshToken(session)
✓ 18 tests passed
✦ committed: fix(auth): sign sessions properly
0 bytes to the cloud
Airplane mode approved ✈
01 — Install

Install the app. Choose a verified model.

The app offers only Laro tiers that have passed their individual release gates, then recommends a fit for your memory. No hosted inference API key.

02 — Understand

Veylaro reads your entire codebase. Locally.

It indexes your repo on-device and builds a live map of your architecture. Your proprietary code is never uploaded, embedded, or logged anywhere.

03 — Act

An agent that edits, runs and tests.

Veylaro plans multi-step changes, writes the code, runs your test suite, reads the failures and fixes them — in a loop, until it's done.

04 — Own

Everything stays on your machine.

Inference, project files, diffs and memory stay on your machine. Optional search sends the displayed query, and account services sync plan data. Turn those off and project work remains local.

0
bytes of your code sent to the cloud
∞
usage limits on Pro — run it all day
40+
languages & frameworks understood
100%
of the model lives on your hardware
Capabilities

A coding environment built
for local execution.

Agentic coding

Plans and executes multi-step engineering tasks: refactors, features, bug hunts, migrations — hands on the keyboard, not just advice.

Persistent memory

Veylaro remembers your architecture, conventions and past decisions between sessions. It gets sharper the longer you work together.

Visible reasoning

Watch the plan form in real time. Every decision, every file touched, every command run — fully inspectable, never a black box.

Deep project understanding

On-device indexing builds a live semantic map of your entire repo — hundreds of thousands of lines — so answers come with real context, instantly.

Git-native workflow

Clean, reviewable diffs. Sensible commit messages. Branch-aware edits. Veylaro works the way your team already ships.

Files & images

Drop in screenshots, logs, PDFs or design mocks. Veylaro reads them locally and turns them into working code.

Zero-latency feel

No round trip to a data center. Tokens stream straight off your GPU — the fastest network request is the one you never make.

Local API access

A localhost API endpoint on Pro. Point your scripts, editors and tools at your own machine and build on top of Veylaro.

Benchmarks

Measured code generation.
Local execution.

Release scores belong to exact checkpoints, runtimes, and harnesses. Historical research runs stay visible, but current tiers remain pending until reproduced.

Historical runs are preserved below with their recorded model identities. They are not scores for the current downloadable GGUF bundles.

Historical HumanEval · 164 tasks · one completion per task
mlx-community/gemma-3-text-4b-it-4bit (base)113/164 passed · per-task report · historical
68.9%
gemma4:12b via Ollama149/164 passed · per-task report · historical
90.9%
Current downloadable models · qualification pending
Laro lite: unsloth/gemma-3-4b-it-GGUF · Q4_K_M. historical MLX run exists; equivalence to this GGUF/runtime has not been established.
Laro med: unsloth/gemma-3-12b-it-GGUF · Q4_K_M. no complete release-attributed standardized score.
Laro max: unsloth/gemma-3-27b-it-GGUF · Q4_K_M. no complete release-attributed standardized score; 24 GB minimum, 32 GB recommended.
Internal project checks · August 2026Recorded checks include 40/41 and 38/41 across four small projects, 16/16 CRM DOM checks, and one 15/15 snake-game app/harness comparison. Several project runs reached their 1,500-second limit.

Passing individual checks is not proof of prompt completion, production-ready tenant security, or parity across every app task. These small internal courses are not SWE-bench or Terminal-Bench scores.

Frontier comparison: no complete, release-attributed SWE-bench or Terminal-Bench result is available for these bundles.
Systems improvements: simulations are design hypotheses, not measured capability gains. Improvements require executed, controlled comparisons.

Model hashes, historical report hashes, and evidence limitations are in the evidence manifest.

Full benchmark breakdown
Private by physics

Keep model inference local.
Choose when the network is used.

Veylaro runs model inference, project indexing, memory, and transcripts on your machine. Optional account, update, and live-search services use the network only when disclosed; local project context is not attached to those requests by default.

Why local wins
VeylaroCloud AI
Your code leaves your machineNeverEvery request
Usage limitsNoneHourly / weekly caps
Works offlineYesNo
Service-side training on your dataNoPolicy-dependent
Cost at heavy usageFlatEscalating

Own your AI.

Veylaro Code puts local models, scoped tools and verified project memory on your own hardware.

Free tier included·Pro from $29/mo·Billed by Stripe · USD & NZD