BeansAI
ModelsPricingDocs
Sign inSign up
ModelsPricingDocs
Sign inSign up

Quickstart

  • Overview

Clients

  • Claude Code
  • CC Switch
  • OpenClaw
  • Roo Code
  • OpenCode
  • Codex CLI
  • GPT Image 2
  • Seedance 2.0
  • SkyReels V4
  • Mureka Song
  • Cursor
  • Cherry Studio

Reference

  • Raw API
← Back to Docs

Roo Code

VS Code AI coding plugin. Connect to BeansAI as a custom Anthropic provider in minutes.

Overview

Roo Code is an AI coding assistant that runs inside VS Code (and VS Code forks like Cursor). It supports plugging in any Anthropic-compatible API endpoint via its graphical settings UI — no config files to edit.

Installation

  1. Open VS Code. Press Ctrl+Shift+X to open the Extensions panel.
  2. Search for Roo Code and click Install.
  3. Once installed, the Roo Code icon appears in the sidebar.

Configuration

  1. Click the Roo Code icon in the sidebar to open its configuration panel.
  2. Under API Provider, select Anthropic.
  3. Paste your BeansAI key (starts with sk-beans-) into API Key.
  4. Check Use custom base URL and enter the BeansAI endpoint.
  5. Select your model (e.g. claude-opus-4-7) and save.
Roo Code settings reference
API Provider:      Anthropic
API Key:           sk-beans-...
Use custom base URL: ✅ (check)
Base URL:          https://api.beansai.dev
Model:             claude-opus-4-7

⚠ Do NOT check:
  "Send Anthropic API key as Anthropic header"
  "Enable 1M context"

⚠ Important: Do not check Send Anthropic API key as Anthropic header or Enable 1M context. Either option breaks request compatibility with BeansAI.

Basic usage

Open any file in VS Code, then click the Roo Code icon. Type your request in the chat box. Roo Code automatically includes the current file as context — select a code block first to scope the change to that snippet.

Tips

  • Select code before typing your prompt — Roo Code will scope the edit to that selection and produce more precise diffs.
  • Roo Code supports multi-file edits in a single conversation. Describe the change in one message and let it touch all relevant files.
  • If a response isn't quite right, click Retry with an amended prompt rather than starting a new conversation — context is preserved.
  • Switch models mid-conversation by changing the model selector in the panel. Useful for drafting with a cheaper model, then refining with Opus.