[opensourcedex]

auto-dev is a free, open source ides & code editors project written in Kotlin and released under MPL-2.0. It has 4,548 GitHub stars, 502 forks and 4 open issues, and was last pushed 6 days ago. On this registry it ranks #23 of 30 tracked projects in IDEs & Code Editors, with 5 head-to-head comparisons available.

What is auto-dev?

AutoDev is an AI-native, multi-agent development platform built on Kotlin Multiplatform that lets developers run coding, review, and document workflows from IntelliJ IDEA, VS Code, the CLI, desktop, mobile, or the web.

What it is

AutoDev Xiuper is a multi-agent coding development platform written in Kotlin and built on Kotlin Multiplatform, published under the MPL-2.0 licence. It ships as an IntelliJ IDEA plugin, a VS Code extension, a CLI tool, a desktop and Android application, an iOS shell, a JS/WASM web app, and a Ktor-based server, all sharing one agent runtime through modules such as mpp-core, mpp-ui, and mpp-server.

The project replaces a fragmented, IDE-only assistant workflow with a single cross-platform runtime: the same agent engine, tooling, and configuration follow a developer from editor to server. Concretely it addresses document research, coding, code review, data query, artifact generation, and web interaction in one place, while discovering AGENTS.md rules from the project hierarchy and injecting them into prompts so the agent respects project-specific conventions.

Key capabilities

  • Core agents implemented in code: DocumentAgent, CodingAgent, CodeReviewAgent, ChatDBAgent, and ArtifactAgent.
  • Tool-driven coding with built-in file system access, grep/glob, shell, web fetch, MCP tools, and renderer integration.
  • SubAgent architecture covering analysis, error recovery, NanoDSL, chart/plot, codebase investigation, domain dictionary, SQL revision, and web agent flows.
  • IDE workflow features including PR review, pre-push review, diagram-related review flows, chat history, token usage tracking, and model/tool configuration UIs.
  • Agent ecosystem support for MCP, A2A agent commands, Claude Skill loading, and SpecKit command integration.
  • Multi-LLM support across OpenAI, Anthropic, Google, DeepSeek, and Ollama.
  • Code intelligence via Tree-sitter parsing for Java, Kotlin, Python, JavaScript/TypeScript, Go, Rust, and C#.

Who uses it and how

  • IntelliJ IDEA users install the plugin from the JetBrains Marketplace to run PR review and pre-push review flows directly inside their editor.
  • VS Code developers install the Phodal.autodev extension from the Visual Studio Marketplace, backed by mpp-core JS bindings.
  • Server teams deploy the mpp-server module, a Ktor-based remote coding agent server, and drive agents over MCP and A2A protocols.
  • Command-line users install the CLI tool globally with npm install -g @xiuper/cli for headless runs.
  • Cross-platform adopters run the Desktop JVM and Android builds from GitHub Releases, or build the iOS target from source.

Getting started

Install the IntelliJ IDEA plugin from the JetBrains Marketplace, the VS Code extension from the Visual Studio Marketplace, or the CLI via npm install -g @xiuper/cli; Desktop and Android builds are distributed through GitHub Releases, and the web app is available at web.xiuper.com.

How it compares

AutoDev does not name specific paid competitors in its documentation, so it stands alongside other AI coding assistants as the only project in this registry built as a Kotlin Multiplatform runtime spanning IDE plugins, CLI, desktop, mobile, web, and server from a single codebase.

When to use it — and when not to

Because one Gradle build covers JVM, Android, iOS, JS, and WASM targets, a self-hoster should expect to operate the Ktor mpp-server and configure model access and tool integrations such as MCP themselves. Note that mpp-vscode, mpp-ios, and mpp-web are present in the repository but excluded from the default root settings.gradle.kts build graph, and the Xiuper edition is labelled Alpha, while AutoDev 2.0 remains the stable branch for those who need a released version.

project readme (upstream, from github) — read inline

AutoDev 3.0 Xiuper (Alpha)

One Platform. All Phases. Every Device.
统一平台 · 全开发阶段 · 跨全设备

AutoDev Xiuper is an AI-native, multi-agent development platform built on Kotlin Multiplatform. Based on the current codebase, it focuses on document research, coding, code review, data query, artifact generation, and web interaction workflows, and targets IntelliJ IDEA, VS Code, CLI, Desktop JVM, Android, iOS, JS/WASM Web, and Server runtimes.

ScreenShot

One Platform Architecture

AutoDev Xiuper One Platform Architecture

Get Started

Download AutoDev Xiuper

Previous Versions

Modules

Module Platform Current Status Description
mpp-core Shared KMP runtime ✅ In default build Shared agent engine, tools, MCP integration, AGENTS.md loading
mpp-ui Desktop / Android / iOS / JS / WASM ✅ In default build Main multiplatform UI surface and app entrypoints
mpp-server JVM server ✅ In default build Ktor-based remote coding agent server
mpp-idea IntelliJ IDEA ✅ Composite build IDEA plugin with dedicated renderers and toolwindow integration
mpp-vscode VS Code ✅ Standalone package in repo VS Code extension backed by mpp-core JS bindings
mpp-viewer / mpp-viewer-web Shared viewer stack ✅ In default build Diagram/viewer support used by UI surfaces
xiuper-ui / xiuper-fs / xiuper-e2e Shared support modules ✅ In default build UI foundation, cross-platform file system, and E2E support
mpp-ios iOS app shell 🔄 Present in repo Native iOS wrapper and build scripts, not included in current root Gradle graph
mpp-web Landing / demo web 🔄 Present in repo Vite-based landing page and demo web entry, not included in current root Gradle graph

Current Status Snapshot

  • The default root Gradle build currently includes mpp-core, mpp-ui, mpp-codegraph, mpp-server, mpp-viewer, mpp-viewer-web, xiuper-ui, xiuper-fs, and xiuper-e2e, with mpp-idea as a composite build.
  • mpp-ui is the main cross-platform application module and already declares JVM, Android, iOS, JS, and WASM targets.
  • mpp-vscode, mpp-ios, and mpp-web are maintained in the repository, but they are not part of the current root settings.gradle.kts default build graph.

Key Features

Xiuper Edition currently exposes these code-backed capabilities:

  • Unified KMP Runtime: Shared agent/runtime code across JVM, Android, iOS, JS, and WASM targets
  • Core Agents in Code: DocumentAgent, CodingAgent, CodeReviewAgent, ChatDBAgent, and ArtifactAgent
  • Tool-Driven Coding Agent: Built-in file system, grep/glob, shell, web fetch, MCP tools, and renderer integration
  • Project Rule Awareness: Automatic AGENTS.md discovery and prompt injection from project hierarchy
  • SubAgent Architecture: Analysis, error recovery, NanoDSL, chart/plot, codebase investigation, domain dictionary, SQL revision, and web agent flows
  • IDE Workflow Features: PR review, pre-push review, diagram-related review flows, chat history, token usage, and model/tool configuration UIs
  • Agent Ecosystem Support: MCP, A2A agent commands, Claude Skill loading, and SpecKit command integration
  • Multi-LLM Support: OpenAI, Anthropic, Google, DeepSeek, Ollama, and more
  • Code Intelligence: Tree-sitter based parsing for Java, Kotlin, Python, JavaScript/TypeScript, Go, Rust, C#
  • Global Ready: Full internationalization (Chinese/English)

Built-in Agents

AutoDev Xiuper currently has the following code-backed top-level agents and adjacent agent capabilities:

Agent Area Description Capabilities Status
Document / Knowledge Requirements / Research DocumentAgent for document querying and feature tree generation DocQL / analysis sub-agent / feature tree ✅ Code-backed
Coding Development CodingAgent with workspace tools and orchestration File system / shell / MCP / sub-agents / AGENTS.md ✅ Code-backed
Review Code Review CodeReviewAgent for review, lint summary, and fix generation Linter / diff review / PR review services ✅ Code-backed
ChatDB Data ChatDBAgent for natural-language database interaction Schema linking / SQL generation / SQL revision ✅ Code-backed
Artifact Rapid Prototyping ArtifactAgent for self-contained runnable outputs HTML / React / Node.js / Python / SVG / Mermaid ✅ Code-backed
Web Agent / WebEdit Web Interaction Web interaction capability exists as web-agent and WebEdit UI flows Page inspection / DOM context / chat-assisted actions 🔄 Experimental

Testing-related capability currently exists mainly through xiuper-e2e and E2E-oriented agent/tooling work, but it is not yet represented as a clearly separated top-level agent in the current code layout.

SubAgents

SubAgents are specialized micro-agents invoked by the main Coding Agent for focused tasks. They follow the "agent as tool" architecture pattern:

SubAgent Purpose Key Features Platform Support
NanoDSL Agent Generate AI-native UI code from natural language descriptions Token-efficient DSL / Component generation / State management / HTTP requests All platforms
PlotDSL Agent Generate statistical charts and data visualizations from natural language ggplot2-inspired syntax / Multiple chart types / Themes / Lets-Plot rendering JVM Desktop & Android
Chart Agent Generate chart configurations for ComposeCharts library Pie/Line/Column/Row charts / Data analysis / Cross-platform rendering All platforms
Analysis Agent Intelligently analyze and summarize any type of content (logs, errors, JSON, code, etc.) Content type detection / Smart summarization / Metadata extraction All platforms
Codebase Investigator Investigate codebase structure, patterns, dependencies, and architectural issues Architecture analysis / Pattern detection / Dependency mapping / Issue identification All platforms
Domain Dict Agent Generate domain dictionaries from codebase analysis for better context understanding Hot file detection / Class/method extraction / Domain term identification All platforms
Error Recovery Agent Analyze errors and suggest fixes with self-healing capabilities Error pattern recognition / Fix suggestion / Auto-retry logic All platforms
SQL Revise Agent Revise and optimize SQL queries based on schema and execution feedback Schema-aware correction / Query optimization / Syntax validation All platforms
E2E Testing Agent Perform end-to-end testing with visual understanding and self-healing locators Natural language test scenario generation / Multi-modal perception / Self-healing All platforms

SubAgents enable modular, composable workflows: complex work is decomposed into focused sub-tasks, each handled by a specialized agent.

License

This code is distributed under the MPL 2.0 license. See LICENSE in this directory.

Frequently asked questions

Is auto-dev free to use?

auto-dev is open source under the MPL-2.0 licence. There is no licence fee and no seat count — you can self-host it or, where the project offers one, pay a vendor for a managed version instead.

What does auto-dev do?

🧙‍AutoDev: the Multi-Agent coding development platform built on Kotlin Multiplatform.

What is auto-dev written in?

auto-dev is primarily written in Kotlin. Its source is publicly available at https://github.com/phodal/auto-dev, and it has 4,548 GitHub stars.