A DirectLink-powered AI coding assistant for VS Code and MTCode Studio β quick to try, self-hosted, and useful for learning, coursework, and small projects.
MTCode AI Phoenix is an AI coding assistant extension for VS Code and MTCode Studio. It provides inline code completion and a sidebar chat interface connected to an LLM server you control.
It gives students, beginners, and small teams an accessible path to self-hosted AI coding assistance. Run MTCode-LLM through MTCode Server, invite users if needed, and connect from the editor without exposing the model server to the public internet.
With a mid-size coding model running on a consumer GPU, AI Phoenix can answer programming questions, explain code, generate examples, and provide lightweight completions. It is well suited for learning, coursework, prototypes, and small software projects, but it is not positioned as a replacement for frontier cloud coding systems on large codebases.
AI Phoenix includes the MTCode Portal core library, so it can connect to DirectLink-published LLM servers without a separate MTCode Portal installation.
AI Phoenix connects your editor to a private coding-assistant model server through MTCode DirectLink.
Get code suggestions as ghost text while you type. Accept the full suggestion, a line, or a word when it is useful.
Ask programming questions, request explanations, generate code, and discuss the current file with Markdown and syntax-highlighted responses.
AI Phoenix talks to self-hosted LLM server directly. MTCode coordinates connection setup, but application data is not relayed through MTCode infrastructure.
See inline completions, chat responses, context handling, and server connection setup in a real coding session. Click to play.
Install and launch MTCode Server, choose MTCode-LLM, and load a coding-assistant model that fits your GPU.
If other students, classmates, or team members should use the server, invite them from your administrator account. Each user gets their own account.
Install the AI Phoenix extension in VS Code, or use MTCode Studio, where it is already included.
Sign in from the AI Phoenix sidebar, choose an authorized LLM server, and wait for the status to show that the server is ready.
Ask questions in the sidebar chat or type in the editor to receive inline completions. The current file can be used as context, and additional files can be attached when needed.
AI Phoenix supports context-aware completion using coding models that understand text before and after the cursor. This is useful for small functions, boilerplate, examples, refactoring assistance, and learning unfamiliar APIs.
The sidebar chat is designed for practical programming help, including explanations, examples, debugging ideas, code generation, and questions about the current file.
AI Phoenix works best with coding-assistant models served through MTCode-LLM or another OpenAI-compatible endpoint published through MTCode Server. Larger models generally provide stronger answers but require more GPU memory and may run more slowly.
We recommend Qwen Coder models as a good starting point because they have been tested extensively with AI Phoenix and work well for both coding chat and inline completion.
MTCode AI Phoenix is included with MTCode Studio. This gives new users a quick path: install Studio, connect AI Phoenix to a DirectLink-published MTCode-LLM server, and start experimenting with AI-assisted coding in one environment.