Frequently asked questions

NodeTool is an open-source agent-first creative workspace. Create and edit images, video, audio, and text with agents that work alongside you. Describe what you want, let the agent build it, then take over whenever you like. You get an editable project, not just a finished file: your workflows, assets, and edits stay together. Use supported local models or connect cloud providers with your own keys.

Yes. The whole project is AGPL-3.0 on GitHub. Studio and Cloud are built from the same source, and nothing is held back for a paid tier. You can host it yourself at any time.

NodeTool Studio stores the workspace, workflows, files, and configured credentials on your machine. Supported open-weight models can run locally through MLX, Ollama, llama.cpp, vLLM, and LM Studio. When you call a hosted model, the request leaves your machine for that provider. Cloud uses hosted project storage; review the current privacy policy and terms before using sensitive material.

The node canvas, sketch pad, storyboard, video timeline, script editor, 3D scene, game editor, and app builder expose operations that agents can call as tools. The agent can change the workflow or project on the same surfaces you use, and the execution record shows tool calls, results, errors, and interventions. The toolbelt is also exposed over MCP for compatible external agents.

Yes. NodeTool runs as an MCP server. Run npx -y --package=@nodetool-ai/cli nodetool mcp install and restart the agent. The command registers NodeTool with Claude Code, Codex, and OpenCode, and nodetool mcp config prints a config block for Cursor, Claude Desktop, and other clients. The agent gets workflows, image, video, and audio generation, and your asset library as tools. Studio is optional. When Studio runs, the agent's changes appear in the open editor.

It means you connect your own provider accounts and pay each provider directly at its published price. NodeTool does not add a credit system or a provider-price markup. The provider still receives the requests you send to its models. You will sometimes see this written as BYOK.

NodeTool Studio, the desktop app, is free and open source. Cloud pricing and access can change during alpha, so check the Cloud page before use. In both editions, provider charges are separate: you connect your own provider account and pay that provider directly. Supported local models in Studio have no provider charge per call, but they use your hardware and electricity.

No. There are no credits and no markup. You bring your own API keys and pay each provider their list price directly. You can also run local models with Ollama, MLX, or llama.cpp in the desktop app for no per-call cost at all.

Use Studio for production work. It is the desktop edition for macOS, Windows, and Linux, and it can use hosted providers without a local GPU or run supported local models on suitable hardware. Cloud is the browser edition for evaluation and lightweight access while it is in alpha. Current import and export coverage can change during the alpha, so verify the project path you need before relying on a cross-edition handoff.

Cloud is for evaluation and lightweight access while the hosted edition is in alpha. Enabled models, storage behavior, import and export coverage, and availability can change. Check the Cloud page, the in-product notices, and the current terms before depending on a particular capability. Use Studio for paid production work.

Cloud uses hosted project storage. The current privacy policy and terms are the source for retention and access rules. If material must remain on your device, use Studio and avoid remote providers for the affected steps.

Studio can keep working without a network when the project uses local files and supported local models already installed on the machine. Remote providers, model downloads, online documentation, and other hosted services still require a connection.

The leading models, including Flux, Seedance, Wan, Veo, Kling, Hailuo, Qwen Image, Whisper, ElevenLabs, and Suno, reached through providers such as FAL, KIE, OpenAI, Anthropic, Gemini, Replicate, Together, Groq, Mistral, OpenRouter, and HuggingFace. Models can also run on your own computer through MLX, Ollama, llama.cpp, vLLM, and LM Studio.

Yes. NodeTool Studio runs open-weight models on your own hardware through MLX (Apple Silicon), Ollama, llama.cpp, vLLM, and LM Studio. Nothing leaves your machine unless you call a cloud provider yourself.

No. Studio can use hosted providers, where the model runs on the provider's hardware. A GPU matters only when you choose to run supported open-weight models on your computer. The requirement then depends on the model and runtime; smaller models may run on CPU at lower speed.

ComfyUI is an editor for image models. NodeTool is the studio around it: image, video, music, and words on one canvas, every major model a click away, and editing tools such as masks, inpaint, relight, and layers built in. Both are open source and both work by connecting blocks on a canvas.

Closed tools tie you to a credit system and a hand-picked list of models. NodeTool is open source and runs on your own provider keys. Your workflows, files, and keys belong to you, and you can switch providers the moment a better model appears.

n8n connects business apps and Flowise builds chatbots that answer from your documents; in both, generating an image, a video, or a soundtrack means an HTTP node you configure by hand. NodeTool makes that work native: image, video, music, and text models, agents, and editing tools on one canvas, with an agent that can author the pipeline. NodeTool is AGPL-3.0 open source rather than fair-code, runs as a desktop app, and calls every provider on your own keys.

Yes. Figma Weave (formerly Weavy, acquired by Figma in October 2025) is closed source, hosted only, and billed in its own AI credits. NodeTool covers the same kind of media work as an AGPL-3.0 open-source workspace: image, video, audio, and text on one canvas, run with your own keys at provider prices, as a desktop app, in the browser, or on your own server.

A node-based workflow is one you build by connecting boxes on a canvas instead of writing code. Each box does one thing, such as loading an image, running a model, or cropping a video, and the lines between them carry the result of one step into the next. The whole thing runs from start to finish, so you can watch every step and rearrange it whenever you like.

A diffusion model makes an image, video, or sound by starting from random noise and clearing it away step by step, guided by your prompt, until a clear result appears. Flux, Stable Diffusion, and most image generators work this way. In NodeTool you run them as blocks alongside everything else.

Before the model answers, NodeTool searches your documents and passes the most relevant passages along with your question, so the answer comes from your material rather than the model's general training. The search and document blocks are built in, so you can put this together on the canvas. The technique is often called RAG, short for retrieval-augmented generation.

A planning agent takes a goal, breaks it into steps, picks a model or tool for each step, and works through them in order, adjusting as it goes. In NodeTool an agent is a block on the canvas, so its results feed into the rest of your workflow.