Flux Studio
About
Bridges Flux's image generation capabilities to coding environments, enabling text-to-image, image-to-image, inpainting, and structural control operations directly within IDEs through TypeScript-to-CLI command translation.
Details
- Author
- jmanhype
- Repository
- jmanhype/mcp-flux-studio
- GitHub stars
- 20
- License
- MIT License
- Categories
- Productivity, Developer Tools, Design, AI, Media, Infrastructure
- Tags
- #integration
Jump to
Setting up with Highlight
This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:
- Download and install Highlight from highlightai.com/download
- Navigate to the plugins tab and select "Add Custom Plugin"
-
Configure the plugin with the settings below
Plugin Name
Flux StudioCommand (node, npx, python, etc.)npxArguments-
Argument 1
-y -
Argument 2
@highlight/mcp-server
Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.
-
Argument 1
- Enable "Start Automatically" if you want the plugin to start when Highlight launches
From the repository
| Variable | Required | Description |
|----------|----------|-------------|
| BFL_API_KEY | Yes | Flux API key |
| FLUX_PATH | No | Path to Flux CLI installation (default: /Users/speed/CascadeProjects/flux) |
| VIRTUAL_ENV | No | If set, uses $VIRTUAL_ENV/bin/python instead of python3 |
Cursor: Settings > Features > MCP. Supports stdio and SSE.
Windsurf/Codeium: Edit ~/.codeium/windsurf/mcp_config.json.
generate
Generates an image based on the provided prompt. Required parameters: prompt. Optional parameters: model, aspect_ratio, width, height, output.
img2img
Transforms an image based on the provided prompt. Required parameters: image, prompt, name. Optional parameters: model, strength (0-1), width, height, output.
inpaint
Inpaints an image based on the provided prompt. Required parameters: image, prompt. Optional parameters: mask_shape (circle/rectangle), position (center/ground), output.
control
Controls an image using specified type and prompt. Required parameters: type (canny/depth/pose), image, prompt. Optional parameters: steps (1-100), guidance, output.
Claude Desktop / Cursor
Paste into your MCP client config file to install this server.
{
"mcpServers": {
"flux studio": {
"env": {},
"args": [
"-y",
"@highlight/mcp-server"
],
"command": "npx"
}
}
}Linux
{
"env": [],
"args": [
"-y",
"@highlight/mcp-server"
],
"command": "npx"
}Macos
{
"env": [],
"args": [
"-y",
"@highlight/mcp-server"
],
"command": "npx"
}Windows
{
"env": [],
"args": [
"/c",
"npx",
"-y",
"@highlight/mcp-server"
],
"command": "cmd"
}mcp-flux-studio
MCP server that wraps the Flux image generation API. Exposes text-to-image, image-to-image, inpainting, and structural control (canny/depth/pose) as MCP tools over stdio. The server itself is TypeScript; it shells out to a Python CLI (fluxcli.py) for actual API calls.
What It Does
Receives MCP tool calls, builds command-line arguments, spawns python3 fluxcli.py <subcommand> ... against a local Flux installation, and returns the output. Requires a BFL_API_KEY for the Flux API and a local copy of the Flux CLI.
Status
| Area | State |
|------|-------|
| MCP transport | stdio |
| Language | TypeScript (server) + Python (CLI wrapper) |
| Flux models | flux.1.1-pro, flux.1-pro, flux.1-dev, flux.1.1-ultra |
| Tests | Jest, 2 test files |
| IDE tested | Cursor v0.45.7+, Windsurf/Codeium Wave 3+ |
| npm package | flux-mcp-server v1.0.0 |
| License | MIT |
MCP Tools
| Tool | Required Params | Optional Params | Output |
|------|----------------|-----------------|--------|
| generate | prompt | model, aspect_ratio, width, height, output | Generated image path |
| img2img | image, prompt, name | model, strength (0-1), width, height, output | Transformed image path |
| inpaint | image, prompt | mask_shape (circle/rectangle), position (center/ground), output | Inpainted image path |
| control | type (canny/depth/pose), image, prompt | steps (1-100), guidance, output | Controlled image path |
Width and height are validated to 256-2048 range.
Setup
Via Smithery
npx -y @smithery/cli install @jmanhype/mcp-flux-studio --client claude
Manual
git clone https://github.com/jmanhype/mcp-flux-studio.git
cd mcp-flux-studio
npm install
npm run build
npm start
Environment Variables
| Variable | Required | Description |
|----------|----------|-------------|
| BFL_API_KEY | Yes | Flux API key |
| FLUX_PATH | No | Path to Flux CLI installation (default: /Users/speed/CascadeProjects/flux) |
| VIRTUAL_ENV | No | If set, uses $VIRTUAL_ENV/bin/python instead of python3 |
IDE Configuration
Cursor: Settings > Features > MCP. Supports stdio and SSE.
Windsurf/Codeium: Edit ~/.codeium/windsurf/mcp_config.json.
Architecture
src/
index.ts — MCP server, tool handlers, Python process spawning
types.ts — TypeScript interfaces for tool arguments
cli/
fluxcli.py — Python CLI that calls the Flux API (not in this repo's src)
tests/
server.test.ts
types.test.ts
Limitations
- Shells out to Python for every tool call; each call spawns a new process
- The default FLUX_PATH is hardcoded to a local directory
- No connection pooling or request queuing for the Flux API
- No image previews returned in MCP responses — only file paths
- The ControlType type is referenced but not imported in index.ts
- No progress reporting during generation
Dependencies
| Package | Version | Purpose |
|---------|---------|---------|
| @modelcontextprotocol/sdk | ^0.1.0 | MCP server protocol |
| dotenv | ^16.0.3 | Environment variable loading |
| typescript | ^5.0.3 | Build toolchain |
| jest | ^29.5.0 | Test runner |
License
MIT
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
