Frequently Asked Questions

CreativeClaw is the creative studio inside Claude. It is a remote MCP server that gives Claude access to every top AI media generation model - image, video, audio, and 3D - through a single connection. Think of it as the Adobe Creative Cloud for AI agents. Instead of switching between separate AI tools, you connect one URL and your AI assistant gains full creative capabilities.

MCP (Model Context Protocol) is an open standard created by Anthropic that lets AI assistants like Claude connect to external tools and services. It works like a plugin system - you give Claude a server URL, and it can use the tools that server provides. CreativeClaw is a remote MCP server, which means there is nothing to install locally. You add one URL to Claude and it gains access to all creative tools instantly.

You add the CreativeClaw MCP URL to your Claude settings. In Claude Desktop, go to Settings > MCP Servers and add the URL. In Claude Code, run the MCP configuration command. The entire setup takes under a minute - no downloads, no local installation, no API keys to manage. Once connected, Claude automatically sees all available creative tools.

CreativeClaw works with any Claude plan that supports MCP connections, including Claude Desktop (Pro and Team plans) and Claude Code. Your Claude subscription and CreativeClaw credits are completely separate - Claude handles the AI conversation, CreativeClaw handles the media generation.

CreativeClaw works with any MCP-compatible client. This includes Claude Desktop, Claude Code, OpenClaw, and any other application that supports the Model Context Protocol. If your AI tool speaks MCP, it can use CreativeClaw.

CreativeClaw supports four categories of media generation. Images: text-to-image generation, image editing, style transfer, and background removal. Video: text-to-video and image-to-video generation, plus video background removal. Audio: text-to-speech with hundreds of voices, emotion control, and multi-speaker dialogue. 3D: text-to-3D and image-to-3D model generation in GLB format. All generated media is stored permanently with a shareable URL.

CreativeClaw offers dozens of curated models across Google, fal.ai, Pika, and OpenAI. For images, Nano Banana 2 is the recommended default, alongside Nano Banana Pro, Grok Imagine Pro, Seedream 5, GPT Image 2, FLUX 2 Pro, and Recraft V3. For video, Gemini Omni Flash is the recommended default, alongside Grok Imagine 1.5, FLUX 3, MiniMax H3, Seedance 2.5, LTX 2.3, Veo 3.1, Kling v3, and Hailuo. Speech options include ElevenLabs v3, MiniMax Speech HD, Dia TTS, and more.

Start with Nano Banana 2 for image generation and Gemini Omni Flash for video; they are the recommended defaults for quality, speed, and value. Switch to a specialist when the job calls for it—for example, Recraft V3 for design work, GPT Image 2 for precise text, Seedance or Veo for premium cinematic video, and ElevenLabs v3 for expressive speech. CreativeClaw also includes model discovery so your agent can help choose.

Multi-variant generation lets you create up to 3 versions of the same image in parallel, each using a different AI model or different settings. For example, you can generate the same product shot with Nano Banana 2, Grok Imagine Pro, and Recraft V3 side by side, compare the results, and pick the best one. This gives you studio-quality output without manually tweaking prompts across separate tools.

Yes. CreativeClaw includes image editing tools that let you modify existing images using text prompts. You control the transformation strength from subtle tweaks (0.1) to complete reimagining (1.0). You can also remove backgrounds from any image to get a transparent PNG, which is useful for product photography, marketing materials, and compositing.

Remotion is a React-based framework for creating programmatic videos. CreativeClaw integrates with Remotion so you can generate dedicated media assets - images, animations, audio - and use them in Remotion video projects. This means you can create fully branded video content with AI-generated visuals, all orchestrated from within Claude.

CreativeClaw uses a simple credit system. $10 gets you 1,000 credits. Each generation costs a specific number of credits depending on the model and output. There are no monthly subscriptions or recurring fees - you buy credits when you need them and they never expire. You only pay for what you generate.

Costs vary by model and output. Nano Banana Lite and Seedream 5 Lite cost 7 credits per image, the recommended Nano Banana 2 costs 16 credits, and Nano Banana Pro costs 30. A five-second video costs 50 credits with Veo 3.1 Lite, 130 with the recommended Gemini Omni Flash, or 400 with Veo 3.1. See the live pricing catalog for resolution tiers, specialist models, and utility costs.

With 1,000 credits you can generate approximately 142 Nano Banana Lite images, 62 Nano Banana 2 images, 33 Nano Banana Pro images, 20 five-second Veo 3.1 Lite videos, 7 five-second Gemini Omni Flash videos, or 2 five-second Veo 3.1 videos. You can mix and match across models and media types.

You can purchase credits directly through CreativeClaw. Credit packs are available in several sizes - 1,000, 3,000, and 7,000 credits. When you run low, Claude can provide a direct link to purchase more credits without leaving your conversation.

All generated media is stored permanently on CreativeClaw's CDN (powered by Cloudflare R2). Every file - image, video, audio, or 3D model - gets a permanent URL that you can share, embed in websites, use in presentations, or post to social media. Your files are always accessible and never expire.

CreativeClaw includes a built-in asset library where all your generated and uploaded media is organized. You can search assets by type (image, video, audio, 3D), by name, tags, or description, and by the prompt that generated them. Every asset can be tagged and described for easy retrieval later. Think of it as your personal creative media library that Claude can access and reference.

CreativeClaw has a theme system for brand management. You save your brand elements - colors, fonts, logos, product images, and style guidelines - as a reusable theme. When generating media, you pass your theme and Claude automatically references your brand DNA. You can create multiple themes for different brands or campaigns, and set a default theme that applies automatically.

Yes. You can upload images, videos, audio, and 3D models to your asset library in three ways: base64 upload for files up to 37MB, URL import for media hosted elsewhere, or interactive drag-and-drop directly in the chat. Uploaded assets get the same permanent URL and organization features as generated ones.

CreativeClaw plugins are standard Claude Code and OpenClaw plugins - a package that bundles the MCP connection, dedicated skills, and common setup into one install. Think of it as the knowledge layer on top of the creative tools you get via the MCP. The plugin knows how to choose the best model for each task, how to write effective prompts for each model, and how to chain tools together into complete creative workflows. It turns Claude from a general assistant with creative tools into a creative professional that knows how to use them.

The plugin includes skills that guide Claude through complex creative workflows end-to-end. This includes skills for choosing the best model for your specific use case, optimizing prompts for each model's strengths, and completing full production workflows - like creating a TikTok product video from scratch, generating an animated video for a scroll-based landing page, or producing a multi-angle product photography sheet. Skills are continuously updated as new models and techniques become available.

Installation takes under a minute. Visit our setup guide at creativeclaw.co/how-to-connect for step-by-step instructions. The short version: you add the CreativeClaw plugin package to your Claude Code or OpenClaw configuration, and everything - MCP connection, skills, and setup - is ready to use immediately.

CreativeClaw plugins currently work with Claude Code (Anthropic's CLI), Claude Desktop (via Cowork), and OpenClaw. Support for Claude on the web (claude.ai) is coming soon. Note that even without the plugin, you can still use CreativeClaw's MCP tools directly in any MCP-compatible client - the plugin just adds the skills and knowledge layer on top.

Yes. CreativeClaw uses Clerk for authentication (including Google OAuth), so we never store your password. Your media files are stored on Cloudflare R2 with per-user isolation - each user's assets are stored in a separate namespace. Generation prompts and metadata are stored in a PostgreSQL database. We do not use your prompts or generated media to train any models. See our privacy policy for full details.

Some media types - especially video and 3D models - take longer to generate (typically 30 seconds to 2 minutes). CreativeClaw handles this with a job queue system. When you request a video or 3D model, you get a job ID immediately. Claude then polls the job status and delivers the result when it is ready. You do not need to wait or keep the conversation open - the job runs in the background on fal.ai's infrastructure.

CreativeClaw reserves or deducts the estimated credits when a generation starts. If it fails because of a server or provider error, those credits are automatically refunded. Failed jobs return a clear error message so your agent can help adjust the prompt or try a different model.