Fattly
The Fattly MCP server lets an AI assistant generate images, videos and audio, upload local images as inputs, and check models/prices/credits through your FATTLY account.
Check credits —
fattly_creditsshows the current balance (free, read-only).List models & prices —
fattly_list_modelsreturns available image/video/audio models with credit costs, optionally filtered by category (free, read-only).Upload a local image —
fattly_upload_imagesends a JPG/PNG/WEBP (max 10 MB) and returns a fal URL to use asinputImageUrl(free).Generate/edit images —
fattly_generate_imagefrom a prompt, or edit with 1–4 uploaded input images (model, aspect ratio, 1–4 outputs); spends credits.Generate videos —
fattly_generate_videotext-to-video or image-to-video (e.g. Kling, Veo, Seedance); waits up to 10 min; spends credits.Generate audio —
fattly_generate_audiofor voiceovers (with voice choice) or music clips (15/30/60 s); spends credits.Safety & billing — SFW prompts only; failed paid generations are refunded automatically.
Note: the README describes 21 tools (including Ad Studio tools like fattly_generate_ad_video), but this schema exposes only the 7 listed above.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Fattly generate an image of a peaceful forest stream"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
fattly — FATTLY CLI + MCP server
Turn a product photo into a UGC video ad, generate AI images, video and voiceovers — from your terminal or straight from AI assistants like Claude, Cursor and VS Code via MCP.
50+ AI models in one account (Veo 3.1, Kling 3, Seedance, Nano Banana, GPT Image, ElevenLabs…) · pay with fattly.app credits that never expire.
MCP (Claude, Cursor, VS Code)
The package is an MCP server — it lets AI assistants generate media through your FATTLY account.
Option A — Hosted (no install, sign in with OAuth)
Connect your client to the hosted server. No npm, no key to paste — you approve access in the browser.
https://fattly.app/api/mcpIn Claude: Settings → Connectors → Add custom connector → paste the URL. Full setup guide: https://fattly.app/claude-mcp
Option B — Local (npx, with an API key)
On fattly.app open Dashboard → API keys and create a key.
Add this to your client's MCP config:
{
"mcpServers": {
"fattly": {
"command": "npx",
"args": ["-y", "fattly", "mcp"],
"env": {
"FATTLY_API_KEY": "gpx_live_YOUR_KEY"
}
}
}
}Installing with an AI agent (Cline, Claude Code, Cursor)? Point it at llms-install.md.
Tools
21 tools — the same set as the hosted server (the local server forwards every call to fattly.app with your key, so new tools appear without updating the package).
Tool | What it does | Spends credits |
Ad Studio | ||
| UGC video ad: a real-looking AI creator presents your product on camera with a real voice (or hands-only POV / unboxing) | yes (~150) |
| Ad image from a product photo: styled product scene or a creator holding it | yes (20) |
| Your own consistent AI creator for ads | yes (20) |
| Ad presets, scenes and creator avatars | no (read-only) |
Images | ||
| Text-to-image and image editing (multi-image edits) | yes |
| Upscale an image | yes |
| Clean cutout without background | yes |
| Repaint a masked area to match a prompt | yes |
| Swap a face (SFW, only faces you have the right to use) | yes |
Video | ||
| Text-to-video and image-to-video (Veo, Kling, Seedance…) | yes |
| Check a video that is still rendering | no (read-only) |
| Bring a still photo to life | yes |
| Trend-style video from a photo and a preset | yes |
| Draw-to-Edit: change a masked region of a video with text | yes |
| Replace the creator in an existing video | yes |
Audio | ||
| Voiceover (text-to-speech) or music | yes |
| Clone a voice (with consent) or reuse a saved one | yes |
| Audio → text with timed segments | yes |
Account | ||
| Current credit balance | no (read-only) |
| Models and their price in credits | no (read-only) |
| Photo (URL, base64 or local file) → URL for edit / ad / image-to-video tools | no |
Every tool carries MCP annotations (readOnlyHint, destructiveHint, idempotentHint), so clients can ask before paid calls. Failed generations are refunded automatically. Content must be safe-for-work.
Related MCP server: mcp-media-engine
CLI
npm install -g fattly # Node.js 18+
fattly login # paste your gpx_live_… keyfattly credits
fattly models video
fattly generate "astronaut cat on Mars" --model nano-banana-2 --out cat.png
fattly video "ocean waves at sunset" --model kling-3-standard --out waves.mp4
fattly audio "Welcome to FATTLY" --voice Rachel --out voice.mp3
fattly audio "calm ambient" --music --seconds 30 --out music.mp3Configuration
Variable | Purpose |
| API key (alternative to |
| API address override (default |
The key saved by fattly login lives in ~/.fattly/config.json with owner-only permissions.
More
Website: https://fattly.app
Even more in the web app: AI dubbing with lip-sync in 5 languages, virtual try-on, Cinema Studio.
Issues: https://github.com/industriesfatty-spec/fattly-mcp/issues
MIT © Fattly
Available Tools
6 toolsfattly_creditsCheck credit balanceARead-onlyIdempotentInspect
Shows the current credit balance on the FATTLY account. Read-only and free — call it before an expensive generation to make sure the user has enough credits.
Inputs: no named parameters.
fattly_generate_audioGenerate a voiceover or musicAInspect
Generates AI audio: a voiceover (mode=speech) from text, or music (mode=music) from a description. Returns a link to the mp3 file. Spends credits (price in fattly_list_models → audio); refunded automatically if the generation fails.
Inputs: mode (required), text (required), voice, seconds.
fattly_generate_imageGenerate an imageAInspect
Generates an AI image from a prompt (or edits uploaded images). Returns links to the finished images. Spends credits: model price × numImages (see fattly_list_models); fails with an error if the balance is too low, and credits are refunded automatically when a generation fails. Prompts must be safe-for-work.
Inputs: prompt (required), model, numImages, aspectRatio, +2 more.
fattly_generate_videoGenerate a videoAInspect
Generates an AI video from a prompt (or from a start image). Takes a few minutes — the tool waits (up to 10 min) and returns a link to the mp4 file. Video is the most expensive operation: it spends the model's credit price (see fattly_list_models), so confirm the model and cost with the user first. Credits are refunded automatically if the generation fails.
Inputs: prompt (required), model, duration, aspectRatio, +1 more.
fattly_list_modelsList models and pricesARead-onlyIdempotentInspect
Returns available models and their cost in credits (images, video, audio). Read-only and free. Use before generating to pick the right model id and to tell the user what a generation will cost.
Inputs: category.
fattly_upload_imageUpload a local imageAInspect
Uploads a local image file (JPG, PNG or WEBP) to FATTLY and returns a public fal URL. Use this to turn a photo on disk into an inputImageUrl for editing / face-swap / image-to-video models: pass the returned URL as inputImageUrl to fattly_generate_image or fattly_generate_video. No credits are charged for the upload.
Inputs: filePath (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 6 tools
Each tool targets a distinct resource/action: credits, models, upload, image generation, video generation, audio generation. No two tools overlap in purpose; the generation tools are clearly separated by output modality.
All tools share the 'fattly_' prefix and use a consistent verb_noun pattern (list_models, upload_image, generate_image). The only minor deviation is 'fattly_credits' which is a noun rather than a verb_noun like 'get_credits' or 'check_credits'.
Six tools is well-scoped for a media generation server: one for account state, one for catalog/pricing, one for input preparation, and three for the core generation modalities. Each tool earns its place.
The core lifecycle is covered: check credits, list models, upload input, generate image/video/audio. Minor gaps exist (no tool to fetch a generation result by ID, no explicit cancel/delete), but the server's stated purpose of generating media is fully covered.
Maintenance
Related MCP Connectors
MCP server for Flux AI image generation
Generate AI images and videos from any compatible MCP client.
MCP server for Wan AI video generation
MCP server for OpenAI API (chat completions, image generation, embeddings) via AceDataCloud
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceAn MCP server to use the fal.ai APIs to generate images and videos.32 PyPI4MIT
- AlicenseAqualityCmaintenanceMCP server for AI-powered image, audio, and video generation, enabling media creation directly from Claude, Cursor, and other MCP clients.1144 npmMIT
- AlicenseBqualityDmaintenanceMCP server for generating and editing images using OpenAI, and creating videos using OpenAI Sora and Google Veo. Enables fetching media from URLs or disk with smart output placement.1416 npm9MIT
- FlicenseNot gradedqualityDmaintenanceAn MCP server for AI-powered image processing (generate, edit, vary, analyze) supporting OpenAI, Gemini, Ideogram, and custom relay endpoints.-