A Claude Code-style chat sidebar for VS Code, powered by Aider and your own LLM provider. Bring your own key — Aider Studio is the experience layer; you choose the model.

Heads-up: This extension orchestrates Aider, which runs either in Docker (default, recommended) or natively. You need one of those installed — see Requirements. It also sends file summaries to whichever LLM provider you configure.

Screenshots

Pick a provider, bring your own key	Chat & streamed answers

Requirements

You need either:

Docker Desktop (recommended) — Aider runs in a container, nothing else to install. On first use, build the image once: docker compose build (or run the "Aider Studio: Build Docker Image" command), or it falls back to pulling/using the default image. Enabled by default (aiderStudio.useDocker).
or a local Aider: pip install aider-chat, with aider on your PATH. Set aiderStudio.useDocker to false.

And an API key for at least one provider (see below).

Getting a key (bring your own model)

Aider Studio works with any Aider/LiteLLM model. The quality of answers is the model's job — pick a capable one. Good free, globally-available options:

OpenRouter — openrouter.ai/keys. Works everywhere; strong free models (e.g. Qwen3 Coder). Preconfigured as a provider. Free models come and go — see the current list at openrouter.ai/models?max_price=0.
Gemini — aistudio.google.com. Generous token limits where the free tier is available (regional restrictions apply).
Groq — console.groq.com. Very fast, but small daily/per-minute limits and weaker models.

The default models are free tiers with real limits. For a Claude Code-like experience, use a capable model (DeepSeek V3, GPT-4-class, Claude via your own key).

Usage

Open the Aider Studio view in the Activity Bar.
Pick a provider, paste your key (stored in VS Code secret storage — never in plaintext).
Ask away. Reference a file by name and it's auto-added to context; when Aider edits a file you get an inline accept / undo diff.

Action	Shortcut
Open chat	`Ctrl+Shift+A`
Add current file to context	`Ctrl+Shift+F`
Switch / add provider	dropdown in the chat header

Features

Flat, streamed transcript with live markdown (code blocks + copy)
Repo map (REPO_MAP.md) so the model is aware of every file, cheaply
@file autocomplete, context-files strip, inline diff accept/undo
Run in Docker with automatic fallback; rate-limit aware
Add any provider/model from the UI — no code changes

Adding a custom provider

Use the "＋ Add custom provider" entry in the dropdown, or add to settings.json:

"aiderStudio.providers": {
  "openrouter-claude": {
    "label": "OpenRouter — Claude Haiku",
    "aiderModel": "openrouter/anthropic/claude-3.5-haiku",
    "apiKeyEnv": "OPENROUTER_API_KEY",
    "apiKeySettingKey": "aiderStudio.openrouterApiKey",
    "freetier": false
  }
}

Privacy & what it does in your repo

Your prompts and file contents/summaries are sent to the LLM provider you configure.
Aider is git-native: Aider Studio ensures a git repo exists (git init if needed) and Aider auto-commits its edits (toggle with aiderStudio.autoCommit). Use ↩ Undo / /undo to revert.
It writes REPO_MAP.md and a .aider/ folder to your workspace and auto-adds them to .gitignore.
Keys are stored in VS Code secret storage.

Configuration

aiderStudio.useDocker, aiderStudio.dockerImage, aiderStudio.aiderPath, aiderStudio.autoCommit, aiderStudio.showDiffOnChange, aiderStudio.repoMap.enabled, aiderStudio.repoMap.summaries, aiderStudio.repoMap.summaryBudget.