AVL Code

Part I — Getting Started

This part covers Chapters 1–3: what AVL Code is and who it's for, how to download and run it and meet the system requirements, and account sign-in, shared quota and usage billing. If this is your first time, start here.

1. Introduction

AVL Code is an AI desktop assistant for coding and security analysis developed by the Antiy 澜砥 (Landi) team. It is distributed as a native cross-platform app and works out of the box. It seamlessly fuses "conversational AI assistant" with "real engineer workbench" — the LLM doesn't just answer questions, it directly reads / writes files, runs commands, inspects logs, performs security analysis and drives tasks according to a plan.

1.1 Who it's for

  • Application and systems engineers who want the AI to actually work in their code repository, not just paste snippets in a browser chat.
  • Security analysts / reverse engineers who want a single interface for sample basic analysis, IOC extraction, disassembly, decompilation and rule matching.
  • DevOps and SRE who want the AI to read logs, edit configs, organise scripts — and to do it under controlled scheduling.
  • Team leads / project managers who want their team's AI workflow to be auditable, approvable and collaborative through group chats.

1.2 Core value proposition

  • Zero-config out of the box: a shared quota is provisioned on sign-in; no manual API key.
  • Local first: every session lives on the local machine, never uploaded to a third party; survives offline and restart.
  • GM-cryptography compliant: skill packs use Chinese national-standard signing and interoperate with open-source ecosystems.
  • Companion messaging: drive the desktop assistant from a phone through a WeChat channel, receive approval cards and collaborate as a team.
  • Industry capability: a built-in set of read-only tools for binary security analysis, ready to triage samples on the spot.
  • Workflow-aware: a four-stage workflow — plan / prepare / execute / assess — lets the AI behave like a disciplined colleague rather than a question-and-answer machine.

1.3 What sets it apart

The difference between AVL Code and general-purpose AI chat is: it actually works inside your computer.

  • Not just code snippets — it edits your project files directly and lets you review the diff.
  • Not just command suggestions — it runs the command and pastes the output back into the conversation.
  • Not just plans — under plan mode it tracks a structured plan and todo list and advances step by step.
  • Not just a single machine — companion messaging pushes approvals and progress to your phone and team chat.