llama.vscode

llama.vscode is a coding assistant for developers who want AI suggestions, chat, and an agent inside VS Code. It lets you use models on your own hardware or connect to external providers.

• Code and text completion: Get suggestions as you type and accept a whole suggestion, a line, or a word, with context from open files, edited files, and copied text.
• Chat and coding agent: Discuss code and give Llama Agent tasks with attached files or selected text; the agent works with local models such as gpt-oss 20B.
• Agent tools: Give the agent access to MCP servers available in VS Code, built-in tools, and custom tools that read files or web pages or execute JavaScript.
• Model choice: Choose separate models for completion, chat, embeddings, and tools, save them in reusable groups, and search for models on Hugging Face from the extension.
• Phone access: Reach your VS Code agents through a Telegram bot.

The extension is open source under the MIT license and runs in VS Code on macOS, Windows, and Linux. Local inference uses llama.cpp, and completion requires a model that supports fill-in-the-middle (FIM). It supports CPU-only setups, though the recommended small CPU models produce lower-quality results. Context reuse helps it handle large amounts of text on lower-end hardware.

With local models, inference runs on your machine. External providers such as OpenRouter process requests remotely, and Telegram access depends on that external service.