Can AI Really Supercharge Your IDE? Inside the AutoDev Open‑Source Copilot

Background

AutoDev was created to test two hypotheses: (1) large enterprises will eventually operate at least one private large‑language model, and (2) only end‑to‑end tools can fully leverage AI to improve software quality and productivity. The project is open‑source on GitHub at https://github.com/unit-mesh/auto-dev.

Assistance Modes

Automatic Mode – Structured Code Generation

Triggered from the IDE’s Context Actions with ⌥⏎ (macOS) or Alt+Enter (Windows/Linux). The mode generates code that conforms to a team’s conventions.

Auto CRUD : an interactive agent reads requirements and iteratively creates or modifies controllers. Supports Kotlin and JavaScript.

Auto Test Generation : one‑click creation and execution of unit tests for JavaScript, Kotlin, and Java.

Auto Code Completion : language‑specific behavior—Java respects configured code‑style rules; Kotlin/Java add contextual classes based on parameters and return types; other languages use similarity algorithms similar to GitHub Copilot or JetBrains AI Assistant.

Each automatic action runs within a configurable “auto context” that supplies project‑specific metadata such as field definitions, method signatures, and coding standards.

Companion Mode – Continuous Assistance

Runs inside the AutoDevChat interface and waits for LLM responses. It can generate commit messages, release notes, code explanations, refactoring suggestions, and DDL statements without leaving the IDE.

Chat Mode – Edge‑Case Interaction

A one‑click chat action is exposed via Context Actions, allowing developers to converse directly with the LLM inside the IDE.

LLM Integration

The repository includes a Python example of a custom LLM server that translates model responses into the format expected by AutoDev. Internal deployments of models such as ChatGLM2 have been tested.

Platform Support

AutoDev is implemented as a JetBrains IntelliJ plugin, reflecting the author’s expertise in JetBrains plugin development. A VS Code version is not planned in the short term because VS Code lacks many IDE‑level integration points required for the full feature set.

Conclusion

AutoDev treats LLMs as co‑integrators that automate repetitive development tasks while keeping engineers within their IDE. The project remains open‑source and welcomes contributions via the GitHub repository.