Konveyor AI (Kai) - AI Developer Tools Tool

Overview

Konveyor AI (Kai) is an open-source, migration-focused developer tool that combines Konveyor static code analysis with retrieval-augmented generation (RAG) to produce targeted code-change suggestions for application modernization. Kai narrows LLM context by using analyzer-derived incidents (files and locations that need changes) and augments prompts with “solved examples” drawn from an organization’s past migrations so suggestions align with how the organization already solves problems. ([github.com](https://github.com/konveyor/kai)) Designed as a model-agnostic “bring your own model” system, Kai supports multiple LLM providers and uses an agentic workflow to iteratively refine outputs, validate suggested fixes (for example via Maven compilation checks), and surface diffs in the developer’s IDE (VS Code) for review and application. Kai is targeted at real-world modernization scenarios (Java EE→Quarkus and similar), can apply fixes directly into source files when accepted, and re-runs incremental analysis to confirm incident resolution. The project is community-driven, in active development, and positioned as an early-stage, experimental tool for teams automating migration tasks. ([raw.githubusercontent.com](https://raw.githubusercontent.com/konveyor/kai/main/docs/getting_started.md))

Installation

Install via npm:

java -version    # ensure Java 17+ is installed (required for analyzer tooling).

mvn -v           # ensure Maven 3.9+ is installed (required for Java builds/validation).

npm install -g vsce    # optional: tools to package/manage VS Code extensions (if building the extension).

Download the latest Kai .vsix from the project releases (editor-extensions/releases) and install in VS Code GUI (Extensions > ... > Install from VSIX...).

OR: code --install-extension path/to/konveyor-kai.vsix    # CLI install of the VSCode extension.

Configure an LLM API key in the Kai extension settings (supports OpenAI, Azure OpenAI, Amazon Bedrock, Google Gemini, Ollama, DeepSeek, OpenShift AI).

Key Features

• RAG-driven suggestions: uses static-analysis incidents to scope LLM prompts for precise fixes.
• Solved-example learning: reuses prior migration fixes to shape future suggestions.
• Model-agnostic BYOM: supports OpenAI, Azure, Bedrock, Gemini, Ollama and OpenShift-compatible endpoints.
• VS Code integration: installs as a .vsix extension and surfaces diffs inline for review and apply.
• Agentic refinement: iteratively validates and refines LLM outputs (e.g., Maven compile checks).
• Incremental analysis: accepts changes and re-runs partial analysis to verify resolution.

Community

Kai is an open-source Konveyor project with active documentation, blog posts, demo videos, and a public roadmap; upstream repo shows community activity (issues, PRs, releases) and a published 0.1.0 release on April 1, 2025. The project encourages contributions (CONTRIBUTING.md) and provides guided scenarios and notebooks for experimentation. Because Kai is early-stage and experimental, the maintainers recommend trying evaluation builds and guided scenarios before production use. For repository details, install instructions, and LLM configuration see the project docs and the Konveyor component page. ([konveyor.io](https://konveyor.io/blog/2025/kai-release-announcement/?utm_source=openai))

Key Information

• Category: Developer Tools
• Type: AI Developer Tools Tool