Skip to main content

Setup Integration Assistant

This guide walks you through setting up the Integration Assistant feature in the FluxLoop VSCode extension.

Prerequisites​

1. Python Environment​

Ensure you have Python 3.11 or newer:

python3 --version
# Should show: Python 3.11.0 or higher

2. Install fluxloop-mcp​

Install the FluxLoop MCP server package:

pip install fluxloop-mcp

Verify installation:

fluxloop-mcp --version
# Should show: fluxloop-mcp 0.1.0 (or higher)

3. Build Knowledge Index​

The MCP server requires a knowledge index built from FluxLoop documentation:

# If you cloned the fluxloop repository
cd packages/mcp
./scripts/rebuild_index.sh

# Or use the command directly
fluxloop-mcp rebuild-index

This creates an index at ~/.fluxloop/mcp/index/dev containing:

  • Indexed documentation chunks
  • Framework recipes
  • Search metadata
tip

The index only needs to be built once. It will be automatically updated when you upgrade fluxloop-mcp.

4. Configure OpenAI API Key​

The Flux Agent uses OpenAI's Chat Completions API for generating suggestions.

Option 1: Store in VSCode Settings

  1. Open VSCode Settings (Cmd+, or Ctrl+,)
  2. Search for "fluxloop openai"
  3. Enter your API key in Fluxloop: Openai Api Key

Option 2: Provide on First Use

The extension will prompt for your API key when you first run Flux Agent. You can choose to:

  • Enter the key and save it to Secret Storage (recommended)
  • Enter the key for this session only
  • Cancel and configure later

Option 3: Manual Settings File

Add to .vscode/settings.json:

{
  "fluxloop.openaiApiKey": "sk-...",
  "fluxloop.openaiModel": "gpt-4o-mini"
}
warning

Never commit API keys to version control. Use workspace-local settings or Secret Storage.

Verify Setup​

Open the Integration view in the FluxLoop activity bar:

  1. Check System Status section:

    • ✅ Python Environment should show your Python version
    • ✅ fluxloop-mcp Package should show installed
    • ✅ MCP Index should show ready

  2. Click Connect MCP to verify the connection

  3. Try a knowledge search:

    • Click Knowledge Search
    • Enter: "How do I integrate FastAPI?"
    • Check the FluxLoop output channel for results

If all status indicators are green, you're ready to use Flux Agent!

Configuration Options​

Model Selection​

Choose the OpenAI model for suggestions (default: gpt-4o-mini):

{
  "fluxloop.openaiModel": "gpt-4o"
}

Supported models:

  • gpt-4o-mini (recommended, fast and cost-effective)
  • gpt-4o (higher quality, slower)
  • gpt-4-turbo (legacy)
  • gpt-3.5-turbo (fastest, lower quality)

MCP Index Path​

Override the default index location:

export MCP_INDEX_PATH="/custom/path/to/index"

Or specify in VSCode settings:

{
  "fluxloop.mcpIndexPath": "/custom/path/to/index"
}

Next Steps​