Flux Image Mcp — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Flux Image Mcp (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Generate images directly inside Claude using FLUX.1-schnell — one of the best open-source image models available today.
Claude doesn't generate images natively. This tool bridges that gap: install it once and Claude will be able to create images on demand from any conversation.
This is an MCP server (Model Context Protocol — Anthropic's open standard for extending Claude with external tools). Once installed, Claude detects image requests automatically and calls this server in the background. You just talk to Claude normally.
You: "Generate a photo of a futuristic city at night, 16:9"
Claude: [calls generate_image tool] → saves the image locally and confirmspython3 --versiongit clone https://github.com/ZZtopBR/flux-image-mcp.git
cd flux-image-mcp
python3 -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
pip install -r requirements.txt#### Claude Code CLI
export HF_TOKEN=hf_your_token_here
claude mcp add flux-image-generator -s user \
-- /path/to/flux-image-mcp/venv/bin/python \
/path/to/flux-image-mcp/server.py#### Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"flux-image-generator": {
"command": "/path/to/flux-image-mcp/venv/bin/python",
"args": ["/path/to/flux-image-mcp/server.py"],
"env": { "HF_TOKEN": "hf_your_token_here" }
}
}
}Restart Claude Desktop. A hammer icon confirms the tool is connected.
Just ask Claude naturally:
"Generate an image of an astronaut riding a horse on Mars"
"Create a 16:9 wallpaper of a Japanese temple in autumn"
"Draw a minimalist logo for a coffee shop"| Option | Resolution | Best for |
|---|---|---|
1:1 | 1024 × 1024 | Social media, general use |
16:9 | 1360 × 768 | Wallpapers, YouTube thumbnails |
9:16 | 768 × 1360 | Phone wallpapers, Stories |
4:3 | 1024 × 768 | Presentations, classic photos |
3:4 | 768 × 1024 | Portraits, book covers |
Images are saved to ~/generated_images/.
Works without Claude too:
python generate_image.py "a futuristic city" --size 16:9 --count 2| Problem | Solution |
|---|---|
| Tool not in Claude Code | Run from inside repo folder; check echo $HF_TOKEN |
| No hammer in Desktop | Quit + relaunch; verify absolute paths; check JSON syntax |
HF_TOKEN is not set | Export token or add to Desktop config env block |
| 403 from HuggingFace | Accept the model license on the HF model page |
| Slow generation | 10–30s is normal for the free API |
MIT
Built with [FLUX.1-schnell](https://huggingface.co/black-forest-labs/FLUX.1-schnell) by Black Forest Labs and [MCP](https://modelcontextprotocol.io) by Anthropic.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.