A lean, transparent coding assistant in your sidebar — where you control
every token.
Most AI coding tools quietly inflate every request with huge hidden prompts you
can't see or change — some spend tens of thousands of tokens on a single "hey."
JustCode sends roughly 550 tokens per request, and most of that is a
system prompt you can read and edit. No hidden bloat, no wasted spend — your
context, your rules.
This extension brings the same lean engine as the JustCode
CLI into a VS Code chat sidebar.
Why JustCode
~550 tokens per request — no hidden bloat inflating every call.
A system prompt you can read and edit — no black box.
Bring your own provider & key — OpenAI, Anthropic, OpenRouter, Qwen
(Alibaba Cloud), Ollama, LM Studio, or any OpenAI-compatible endpoint.
See everything it does — live tool activity and inline approvals before
any file is written or command is run.
MCP servers and chat modes (Build / Ask / Plan + custom), sharing the
same engine as the CLI.
Getting started
Install the extension and open the JustCode view from the activity bar.
Connect a provider — either configure one with the
JustCode CLI or set your provider environment
variables. The panel shows a notice with instructions until one is connected.
Pick a model and start chatting.
Features
Streaming chat with live tool activity.
Inline tool approvals — nothing touches your workspace without your OK.
In-tool questions, provider/model selection, and one-click new sessions.
Themed entirely with your VS Code color theme.
Learn more
Full documentation, the always-current list of tools and commands, and the CLI
live at justcodeapp.dev.