Fattly is an AI content generation platform whose core capability is to let users create images, videos, voiceovers, avatars, and UGC ads through prompts. This repository, fattly-mcp, is Fattly's official CLI and MCP client, enabling users to call Fattly's generation features from the terminal or directly within AI assistants such as Claude, Cursor, and VS Code.
In terms of main capabilities, Fattly integrates more than 50 AI models into a single account. The models listed in the materials include Veo 3.1, Kling 3, Seedance, Nano Banana, GPT Image, ElevenLabs, and others. Users can use it to turn product photos into UGC video ads, and can also generate AI images, videos, and voiceovers. Paid usage uses fattly.app credits, and credits do not expire.
In terms of usage, the package itself is an MCP server that allows AI assistants to generate media content through the user's FATTLY account. There are two ways to connect:
Option A - Hosted mode (no installation required, sign in with OAuth): connect the client to the hosted server. No npm is needed, and no key needs to be pasted; simply approve access in the browser. In Claude, the path is: Settings → Connectors → Add custom connector → paste the URL. See the full setup guide at https://fattly.app/claude-mcp。
Option B - Local mode (uses npx, requires an API key): open Dashboard → API keys in fattly.app to create a key, then add the corresponding configuration to the client's MCP configuration. If installing with an AI agent (Cline, Claude Code, Cursor), you can have it refer to llms-install.md.
This MCP server provides 21 tools, the same as the hosted server. The local server forwards each call along with the key to fattly.app, so new tools can appear without updating the package. Each tool carries MCP annotations (readOnlyHint, destructiveHint, idempotentHint), and clients can ask the user before paid calls. Failed generations are automatically refunded. Content must comply with safety standards (safe-for-work). Keys saved via fattly login are stored in ~/.fattly/config.json with owner-only permissions.
Applicable scenarios include: batch or scripted generation of AI images, videos, and voiceovers from the terminal; generating media assets directly in AI assistant conversations; quickly turning product photos into UGC video ads; and creative workflows that require unified access to multiple mainstream generation models.
Limitations and additional notes: the web app also offers more features, such as AI dubbing with lip sync in 5 languages, virtual try-on, and Cinema Studio. Content must be safe and compliant. Information such as specific pricing, detailed parameters for each model, generation duration, and resolution is not provided in the materials. For feedback, visit https://github.com/industriesfatty-spec/fattly-mcp/issues。
