🧙AutoDev: The AI-powered coding wizard with multilingual support 🌐, auto code generation 🏗️, and a helpful bug-slaying assistant 🐞! Customizable prompts 🎨 and a magic Auto Dev/Testing/Document/Agent feature 🧪 included! 🚀
Rebuilt with Kotlin Multiplatform (KMP) to deliver a truly cross-platform AI Coding Agent ecosystem.
Current Versions:
- IntelliJ Plugin:
v2.4.6 - New Intellij Plugin: https://plugins.jetbrains.com/plugin/29223-autodev-experiment
- MPP Modules (Core/UI/Server):
v0.3.4 - VSCode Extension:
v0.5.x
| Module | Platform Support | Description |
|---|---|---|
| mpp-core | JVM, JS, WASM, Android, iOS | AI Agent engine, DevIns compiler, tool system, LLM integration, MCP protocol |
| mpp-codegraph | JVM, JS | TreeSitter-based code parsing & graph building (8+ languages) |
| Module | Platform | Status | Description |
|---|---|---|---|
| mpp-idea | IntelliJ IDEA | ✅ Production | Jewel UI, Agent toolwindow, code review, remote agent |
| mpp-vscode | VSCode | ✅ Production | CodeLens, auto test/doc, MCP protocol, Tree-sitter |
| mpp-ui (Desktop) | macOS/Windows/Linux | ✅ Production | Compose Multiplatform desktop app |
| mpp-ui (CLI) | Terminal (Node.js) | ✅ Production | Terminal UI (React/Ink), local/server mode |
| mpp-ui (Android) | Android | 🚧 In Progress | Native Android app |
| mpp-web (Web) | Web | 🚧 In Progress | Web app |
| mpp-ios | iOS | 🚧 In Progress | Native iOS app (SwiftUI + Compose) |
| Module | Platform | Features |
|---|---|---|
| mpp-server | JVM (Ktor) | HTTP API, SSE streaming, remote project management |
| mpp-viewer | Multiplatform | Universal viewer API (code, Markdown, images, PDF, etc.) |
| mpp-viewer-web | JVM, Android, iOS | WebView implementation, Monaco Editor integration |
- Unified Codebase: Core logic shared across all platforms - write once, run everywhere
- Native Performance: Compiled natively for each platform with zero overhead
- Full AI Agent: Built-in Coding Agent, tool system, multi-LLM support (OpenAI, Anthropic, Google, DeepSeek, Ollama, etc.)
- DevIns Language: Executable AI Agent scripting language
- MCP Protocol: Model Context Protocol support for extensible tool ecosystem
- Code Understanding: TreeSitter-based multi-language parsing (Java, Kotlin, Python, JS, TS, Go, Rust, C#)
- Internationalization: Chinese/English UI support
- Web Demo: https://unit-mesh.github.io/auto-dev/
- VSCode Extension: Visual Studio Marketplace
- CLI Tool:
npm install -g @autodev/cli
Video demo (YouTube) — English
AutoDev Sketch is an IDE canvas feature provided by Shire, designed to simplify interactions and enhance the developer experience within the IDE.
* means requires additional plugin installation.
VSCode Version: https://github.com/unit-mesh/auto-dev-vscode
🆕🆕🆕: New AI agent language: https://github.com/phodal/shire
Video demo (Bilibili) - 中文/Chinese
Here is the AutoDev architecture:
Features:
- Sketch coding Agent
- Agentic drive coding workflow with Sketch viewer.
- Auto development mode
AutoCRUD (Spring framework). With DevTi Protocol (likedevti://story/github/1102) will auto generate Model-Controller-Service-Repository code.- AutoSQL (required Database plugin). Context-aware SQL generation.
- AutoPage (React). Context-aware Web Page generation.
AutoArkUI (HarmonyOS). Auto generate HarmonyOS ArkUI code.- Auto Testing. create unit test intention, auto run unit test and try to fix test.
- Auto Document. Auto generate document.
- Copilot mode
- AutoDev will help you find bug, explain code, trace exception, generate commits, and more.
- Pattern specific. Based on your code context like (Controller, Service
import), AutoDev will suggest the best code to you. - Related code. Based on recent file changes, AutoDev will call calculate similar chunk to generate the best code.
- Chat with AI. Chat with selection code and context-aware code.
- Customize.
- Custom specification of prompt. For example, Controller, Service, Repository, Model, etc.
- Custom intention action. You can add your own intention action.
- Custom LLM Server. You can customize your LLM Server in
Settings->Tools->AutoDev - Custom Living documentation. Customize your own living documentation, like annotation.
- Team AI. Customize your team prompts in codebase, and distribute to your team.
- Prompt override. You can override AutoDev's prompt in your codebase.
- SDLC
- VCS. Generate/improve commit message, release note, and more.
- Code Review. Generate code-review content.
- Smart Refactoring. AI based Rename, refactoring with code smell, refactoring suggetion and more.
- Dockerfile. Based on your project, generate Dockerfile.
- CI/CD config. Based on build tool, generate CI/CD config file, like
.github/workflows/build.yml. - Terminal. In Terminal ToolWindow, you can use custom input to generate shell/command
- Custom AI Agent
- Executable AI Agent language: DevIns.
- Custom AI Agent. You can integrate your own AI Agent into AutoDev.
- Model
AutoDev fine-tune models:
download from HuggingFace
| name | model download (HuggingFace) | model download (OpenBayes) |
|---|---|---|
| DeepSeek 6.7B | AutoDev Coder | AutoDev Coder |
We follow Chapi AST analysis engine for language support tier.
| Features | Java | Python | Go | Kotlin | JS/TS | C/C++ | C# | Scala | Rust |
|---|---|---|---|---|---|---|---|---|---|
| Chat Language Context | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ||
| Structure AST | ✅ | ✅ | ✅ | ✅ | ✅ | ||||
| Doc Generation | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | |||
| Precision Test Generation | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | |||
| Precision Code Generation | ✅ | ✅ |
see in exts
- Copilot-Explorer Hacky repo to see what the Copilot extension sends to the server.
- GitHub Copilot a small part of Copilot Performance logs.
- 花了大半个月,我终于逆向分析了Github Copilot
Welcome to add your company here.
- Thoughtworks, a leading technology consultancy.
@autodev/cliInspired and Tools system based on Google Gemini CLI withApache License 2.0
Inspired by:
- Multiple target inspired by: https://github.com/intellij-rust/intellij-rust
- SimilarFile inspired by: GitHub Copilot
- DevIn Language refs on JetBrains' Markdown Util, which is licensed under the Apache 2.0 license.
- Stream Diff based on Continue Dev under the Apache 2.0 license.
- Ripgrep inspired by Cline under the Apache 2.0 license.
- MCP based on JetBrains' MCP
This code is distributed under the MPL 2.0 license. See LICENSE in this directory.












